<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Iro Boutsi on Medium]]></title>
        <description><![CDATA[Stories by Iro Boutsi on Medium]]></description>
        <link>https://medium.com/@iboutsi?source=rss-cc0277c9a449------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*nDgf17DTJPs0CNjlNk5d0w.jpeg</url>
            <title>Stories by Iro Boutsi on Medium</title>
            <link>https://medium.com/@iboutsi?source=rss-cc0277c9a449------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 21:22:15 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@iboutsi/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[From Images to Meshes: The Production Pipeline Behind 3D Reconstruction]]></title>
            <link>https://medium.com/@iboutsi/from-images-to-meshes-the-production-pipeline-behind-3d-reconstruction-0a57d3e3a74f?source=rss-cc0277c9a449------2</link>
            <guid isPermaLink="false">https://medium.com/p/0a57d3e3a74f</guid>
            <category><![CDATA[3d-modeling]]></category>
            <category><![CDATA[3d-scanner]]></category>
            <category><![CDATA[photogrammetry]]></category>
            <category><![CDATA[computer-vision]]></category>
            <dc:creator><![CDATA[Iro Boutsi]]></dc:creator>
            <pubDate>Mon, 20 Apr 2026 12:55:23 GMT</pubDate>
            <atom:updated>2026-04-20T12:58:04.982Z</atom:updated>
            <content:encoded><![CDATA[<p>Meshes can be generated from images captured by camera sensors, data collected by scanners, or even a hybrid method that integrates both approaches. The choice of method or combination depends on the surface characteristics, the level of accuracy required, cost and the characteristics of the object or scene being modeled. A <em>scene</em> is a collection of geometrically defined objects in 3D object space, with associated lighting and viewing parameters.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fsketchfab.com%2Fmodels%2F4bff096a20064d65b8af65ab8c51d9cf%2Fembed&amp;display_name=Sketchfab&amp;url=https%3A%2F%2Fsketchfab.com%2F3d-models%2Fmar-saba-monastery-4bff096a20064d65b8af65ab8c51d9cf&amp;image=https%3A%2F%2Fmedia.sketchfab.com%2Fmodels%2F4bff096a20064d65b8af65ab8c51d9cf%2Fthumbnails%2Fd91f9ac626104006bbc0946de3f2d8e9%2F3f21de522d5041108bf1238b7cc1bb56.jpeg&amp;type=text%2Fhtml&amp;schema=sketchfab" width="640" height="360" frameborder="0" scrolling="no"><a href="https://medium.com/media/25b6f45669a37c2db96077fd6d4979e5/href">https://medium.com/media/25b6f45669a37c2db96077fd6d4979e5/href</a></iframe><h4>1.1 Image-based techniques</h4><p>The multi-image process is usually referred to as multi-view stereo (MVS) (Furukawa &amp; Hernández, 2015). Assuming that the input dataset comprises overlapping images, the stages are summarized as follows (Verykokou &amp; Ioannidis, 2023):</p><p>i. <strong>Image orientation</strong>: It concerns the calculation of the exterior orientation and (optionally) interior orientation of the images and the generation of a sparse point cloud, which consists of the 3D coordinates of the tie points, e.g., homologous feature points. It is usually performed by applying methods of (i) detection of overlapping images, (ii) image matching and feature tracking, and (iii) structure from motion (SfM). Whether the sparse point cloud is used in the subsequent steps depends on the applied method.</p><p>ii. <strong>Depth map generation</strong>: It is performed via dense image matching for a subset of the overlapping image pairs of known interior and exterior orientation (or for all overlapping image pairs) and produces a set of depth maps for the reference images.</p><p>iii. <strong>Surface reconstruction</strong>: It is assumed that surface generation is incorporated in the 3D reconstruction procedure taking into consideration image orientation to optimize a photo-consistency score over the reconstructed surface. The mesh is generated either directly from the depth maps or indirectly by the estimated dense point cloud. The first case allows for the use of all the information contained in the original images and does not rely on the, often filtered and merged, point cloud. The fusion of depth maps for dense point cloud generation is a middle step followed by some 3D reconstruction algorithms (Merrell et al., 2007; Y. Qiao et al., 2023). Alternatively, the sparse point cloud of SfM is densified without the use of depth maps (Hirschmuller, 2008; Hodlmoser et al., 2013). Some indicative 3D meshing methods are based on triangulated surfaces such as Delaunay triangulation (Cazals &amp; Giesen, 2006; Engwirda, 2014.) and Voronoi diagram (Amenta et al., 1998), volumetric representations such as Poisson reconstruction method (Kazhdan et al., 2006), function fitting approaches etc. The above methods produce a triangular model using all or most of the points, which must be additionally accompanied by normal vectors.</p><p>iv. <strong>Texture mapping stage</strong>: It concerns color encoding introducing an additional per-vertex property, the UV coordinates of texture maps. The generated 2D texture image is mapped in the mesh using the images of known interior and exterior orientation (Z. Chen et al., 2012; Matsushita &amp; Kaneko, 1999; Weinhaus &amp; Devarajan, 1997). For triangular meshes, each triangle is projected in the images in which it should be visible even if partially or fully occluded, connected parts of the same image that are used to texture the triangles of the mesh are incrementally stored (“copy”). Εach vertex of each triangle is assigned texture coordinates, corresponding to row and column numbers of the 2D texture map. For non-textured areas or when texture coordinates fall outside the boundaries of the texture image interpolation is performed.</p><h4>1.2 Scanner-based techniques</h4><p>Scanner-based techniques provide a direct way for capturing 3D geometry from objects in the form of dense point clouds or volumetric representations. Most common technologies are laser scanners that can be distinguished into four basic categories: triangulation scanners, time-of-flight scanners, phase-shift scanners and structured light scanners (Muralikrishnan, 2021). The laser scanner collects color information via a camera sensor or calculates the intensity of the reflected signal. Thus, the produced point cloud is accompanied by color values (usually RGB) or grayscale values, respectively. The captured dense point clouds are initially stored in the local coordinate system of the scanner. When scanning an area from two or more positions, a 3D Euclidean transformation between the point clouds must be computed to align them into a single coordinate system. The problem of aligning two sets of points with one another or merging the point clouds is called <em>registration</em>. Retroreflective targets, when available, can be measured by topographical methods and used to transform the point clouds from the scanner’s local reference system to the ground reference system. However, target availability is not a prerequisite for registration. Point clouds can be aligned using homologous points on overlaps between successive scans. If an initial estimation of their relative rigid-body transform is known, iterative approximation methods, such as iterative closest point — ICP (Y. Chen &amp; Medioni, 1991), feature-based methods, such as the hypothesis-and-test algorithm Random sample consensus (RANSAC) (Fischler &amp; Bolles, 1981) and normal distribution transform (NDT) (Biber &amp; Strasser, 2003) can be employed. The alignment of two point sets can be also, considered as a probability density estimation problem and solved by coherent point drift (CPD) (Myronenko &amp; Xubo Song, 2010). Following merging, the point cloud may undergo further processing, such as outlier removal and denoising. If a mesh is required, an initial surface approximation is generated by the refined point cloud using techniques, such as marching cubes (Newman &amp; Yi, 2006) or point cloud triangulation(P. Su &amp; Scot Drysdale, 1997; Wolff et al., 2016). Texture mapping can be applied to the mesh if images are available, requiring the calculation of interior and exterior orientation parameters using an SfM-MVS pipeline.</p><blockquote>If you would like to cite this work, please use: <strong>Boutsi, A.-M., 2025. Photogrammetric techniques and algorithms optimization for the development of interactive augmented reality systems. NTUA, Athens. </strong><a href="https://doi.org/10.12681/eadd/59886"><strong>https://doi.org/10.12681/eadd/59886</strong></a></blockquote><h4>References</h4><p>Furukawa, Y., &amp; Hernández, C. (2015). Multi-View Stereo: A Tutorial. <em>Foundations and Trends® in Computer Graphics and Vision</em>, <em>9</em>(1–2), 1–148. <a href="https://doi.org/10.1561/0600000052">https://doi.org/10.1561/0600000052</a></p><p>Verykokou, S., &amp; Ioannidis, C. (2023). An Overview on Image-Based and Scanner-Based 3D Modeling Technologies. <em>Sensors</em>, <em>23</em>(2), 596. <a href="https://doi.org/10.3390/s23020596">https://doi.org/10.3390/s23020596</a></p><p>Merrell, P., Akbarzadeh, A., Wang, L., Mordohai, P., Frahm, J.-M., Yang, R., Nister, D., &amp; Pollefeys, M. (2007). Real-Time Visibility-Based Fusion of Depth Maps. <em>2007 IEEE 11th International Conference on Computer Vision</em>, 1–8. <a href="https://doi.org/10.1109/ICCV.2007.4408984">https://doi.org/10.1109/ICCV.2007.4408984</a></p><p>Qiao, Y., Lv, N., &amp; Zhang, S. (2023). 3D Reconstruction and Measurement Analysis of a Dense Point Cloud Fused with a Depth Image. <em>International Journal of Optics</em>, <em>2023</em>, 1–12. <a href="https://doi.org/10.1155/2023/6826981">https://doi.org/10.1155/2023/6826981</a></p><p>Hirschmuller, H. (2008). Stereo Processing by Semiglobal Matching and Mutual Information. <em>IEEE Transactions on Pattern Analysis and Machine Intelligence</em>, <em>30</em>(2), 328–341. <a href="https://doi.org/10.1109/TPAMI.2007.1166">https://doi.org/10.1109/TPAMI.2007.1166</a></p><p>Hodlmoser, M., Micusik, B., &amp; Kampel, M. (2013). Sparse Point Cloud Densification by Combining Multiple Segmentation Methods. <em>2013 International Conference on 3D Vision</em>, 438–445. <a href="https://doi.org/10.1109/3DV.2013.64">https://doi.org/10.1109/3DV.2013.64</a></p><p>Cazals, F., &amp; Giesen, J. (2006). Delaunay Triangulation Based Surface Reconstruction. In J.-D. Boissonnat &amp; M. Teillaud (Eds.), <em>Effective Computational Geometry for Curves and Surfaces</em> (pp. 231–276). Springer Berlin Heidelberg. <a href="https://doi.org/10.1007/978-3-540-33259-6_6">https://doi.org/10.1007/978-3-540-33259-6_6</a></p><p>Engwirda, D., (2014). Locally optimal Delaunay-refinement and optimisation-based mesh generation. <a href="https://ses.library.usyd.edu.au/handle/2123/13148">https://ses.library.usyd.edu.au/handle/2123/13148</a></p><p>Amenta, N., Bern, M., &amp; Kamvysselis, M. (1998). A new Voronoi-based surface reconstruction algorithm. <em>Proceedings of the 25th Annual Conference on Computer Graphics and Interactive Techniques — SIGGRAPH ’98</em>, 415–421. <a href="https://doi.org/10.1145/280814.280947">https://doi.org/10.1145/280814.280947</a></p><p>Kazhdan, M., Bolitho, M., &amp; Hoppe, H. (2006). Poisson Surface Reconstruction. <em>Symposium on Geometry Processing</em>, 10 pages. <a href="https://doi.org/10.2312/SGP/SGP06/061-070">https://doi.org/10.2312/SGP/SGP06/061-070</a></p><p>Chen, Z., Zhou, J., Chen, Y., &amp; Wang, G. (2012). 3D Texture Mapping in Multi-view Reconstruction.<em>Advances in Visual Computing</em> (Vol. 7431, pp. 359–371). Springer Berlin Heidelberg. <a href="https://doi.org/10.1007/978-3-642-33179-4_35">https://doi.org/10.1007/978-3-642-33179-4_35</a></p><p>Matsushita, K., &amp; Kaneko, T. (1999). Efficient and Handy Texture Mapping on 3D Surfaces. <em>Computer Graphics Forum</em>, <em>18</em>(3), 349–358. <a href="https://doi.org/10.1111/1467-8659.00355">https://doi.org/10.1111/1467-8659.00355</a></p><p>Weinhaus, F. M., &amp; Devarajan, V. (1997). Texture mapping 3D models of real-world scenes. <em>ACM Computing Surveys</em>, <em>29</em>(4), 325–365. <a href="https://doi.org/10.1145/267580.267583">https://doi.org/10.1145/267580.267583</a></p><p>Chen, Y., &amp; Medioni, G. (1991). Object modeling by registration of multiple range images. <em>Proceedings. 1991 IEEE International Conference on Robotics and Automation</em>, 2724–2729. <a href="https://doi.org/10.1109/ROBOT.1991.132043">https://doi.org/10.1109/ROBOT.1991.132043</a></p><p>Fischler, M. A., &amp; Bolles, R. C. (1981). Random sample consensus: A paradigm for model fitting with applications to image analysis and automated cartography. <em>Communications of the ACM</em>, <em>24</em>(6), 381–395. <a href="https://doi.org/10.1145/358669.358692">https://doi.org/10.1145/358669.358692</a></p><p>Biber, P., &amp; Strasser, W. (2003). The normal distributions transform: A new approach to laser scan matching. <em>Proceedings 2003 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2003) (Cat. №03CH37453)</em>, <em>3</em>, 2743–2748. <a href="https://doi.org/10.1109/IROS.2003.1249285">https://doi.org/10.1109/IROS.2003.1249285</a></p><p>Myronenko, A. &amp; Xubo Song. (2010). Point Set Registration: Coherent Point Drift. <em>IEEE Transactions on Pattern Analysis and Machine Intelligence</em>, <em>32</em>(12), 2262–2275. <a href="https://doi.org/10.1109/TPAMI.2010.46">https://doi.org/10.1109/TPAMI.2010.46</a></p><p>Newman, T. S., &amp; Yi, H. (2006). A survey of the marching cubes algorithm. <em>Computers &amp; Graphics</em>, <em>30</em>(5), 854–879. <a href="https://doi.org/10.1016/j.cag.2006.07.021">https://doi.org/10.1016/j.cag.2006.07.021</a></p><p>Su, P., &amp; Scot Drysdale, R. L. (1997). A comparison of sequential Delaunay triangulation algorithms. <em>Computational Geometry</em>, <em>7</em>(5–6), 361–385. <a href="https://doi.org/10.1016/S0925-7721(96)00025-9">https://doi.org/10.1016/S0925-7721(96)00025-9</a></p><p>Wolff, K., Kim, C., Zimmer, H., Schroers, C., Botsch, M., Sorkine-Hornung, O., &amp; Sorkine-Hornung, A. (2016). Point Cloud Noise and Outlier Removal for Image-Based 3D Reconstruction. <em>2016 Fourth International Conference on 3D Vision (3DV)</em>, 118–127. <a href="https://doi.org/10.1109/3DV.2016.20">https://doi.org/10.1109/3DV.2016.20</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0a57d3e3a74f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Mesh Structures Explained: Representation, Types, and Connectivity]]></title>
            <link>https://medium.com/@iboutsi/mesh-structures-explained-representation-types-and-connectivity-e625f13cced1?source=rss-cc0277c9a449------2</link>
            <guid isPermaLink="false">https://medium.com/p/e625f13cced1</guid>
            <category><![CDATA[rendering]]></category>
            <category><![CDATA[3d-modeling]]></category>
            <category><![CDATA[photogrammetry]]></category>
            <category><![CDATA[data-mesh]]></category>
            <category><![CDATA[geometry]]></category>
            <dc:creator><![CDATA[Iro Boutsi]]></dc:creator>
            <pubDate>Wed, 08 Apr 2026 09:30:35 GMT</pubDate>
            <atom:updated>2026-04-20T12:59:11.428Z</atom:updated>
            <content:encoded><![CDATA[<p>Meshes are one of the core building blocks of 3D digital geometry, yet their structure can be surprisingly rich once you look beyond the surface. This first part introduces the main concepts behind mesh representation, connectivity, manifold properties and attribute storage, setting the foundation for the rest of the discussion.</p><p><strong>Representation</strong></p><p>A <em>mesh</em> is defined as a collection of vertices (points) connected by faces (polygons) that are enclosed by edges (lines). The vertices are typically defined in 3D space (ℝ³). However, meshes can exist in other dimensions. 2D meshes are confined to a plane, comprising vertices and edges and they are commonly used to describe shapes or boundaries within an image. A 2.5D mesh informally indicates a mesh that can be projected on a plane without self-intersections. Its vertices and edges are defined in 2D coordinates, but the mesh still conveys information about depth or height. In a 4D mesh the coordinates of its vertices belong in ℝ⁴ and represent complex relationships and structures in higher-dimensional geometrical spaces. Very common subclasses of 3D polygonal meshes are <em>triangular meshes</em>, which require all faces to be triangles, <em>tetrahedral meshes</em>, where every face is a tetrahedron and <em>quad meshes</em> where faces consist of quadrilaterals. While quad meshes may offer certain advantages, such as potentially fewer polygons for the same level of detail, their drawbacks in terms of rendering complexity and hardware support make triangles the preferred choice for most 3D applications. Triangular meshes is a long-established data structure in computer graphics and well-suited for hardware optimization on GPUs.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fsketchfab.com%2Fmodels%2F16b189497b6b461590ede3279de9db15%2Fembed&amp;display_name=Sketchfab&amp;url=https%3A%2F%2Fsketchfab.com%2F3d-models%2Fpieta-16b189497b6b461590ede3279de9db15&amp;image=https%3A%2F%2Fmedia.sketchfab.com%2Fmodels%2F16b189497b6b461590ede3279de9db15%2Fthumbnails%2F2b4584b810a7413b90de17a011777fd4%2F7a13c23bfb4040de8663d5eed09d159b.jpeg&amp;type=text%2Fhtml&amp;schema=sketchfab" width="640" height="360" frameborder="0" scrolling="no"><a href="https://medium.com/media/4c173c59c50f08d566c1fb9fb2703e75/href">https://medium.com/media/4c173c59c50f08d566c1fb9fb2703e75/href</a></iframe><h4>Types</h4><p>A <em>conforming</em> mesh is a type of mesh that meets certain geometric criteria: proper edge adjacency and manifold neighborhood. At most two faces should share a common edge to prevent overlaps and self-intersections. If additionally, the neighborhood of each vertex is homeomorphic to a closed ball, the mesh is said to be <em>manifold</em>. Any edge may be shared by either one or two incident faces and the set of faces connected to each vertex forms a single fan of triangles. An edge with two incident faces is said to be <em>internal</em>, while an edge with just one incident face is said to be boundary. It is also referred to as a compact surface. The <em>valence</em> of a vertex is the number of its incident edges, while its <em>star</em> is the set of its incident faces and edges. In a triangular mesh, a vertex is considered <em>regular</em> if it has valence 6. In a <em>regular</em> mesh, every of its vertices is regular while in a <em>semi-regular</em> mesh only a few vertices are <em>extraordinary</em>, i.e. they have more or fewer edges than the regular vertices and are typically introduced at the boundaries of the mesh. Extraordinary vertices can also be called <em>irregular</em> vertices or <em>non-conforming</em> vertices. Along with its position, each vertex can store additional attributes such as color, a normal vector and texture coordinates. Encoding mesh connectivity and editing meshes is much simpler for regular or semi-regular meshes, though, in general, regular meshes require much higher resolution to reach the same accuracy (Ponchio, 2008). Their hierarchical structure makes them ideal for multi-resolution analysis (coarsification by downsampling and smoothing) and syndissertation (subvision and adding of details) (Alliez et al., 2008).</p><h4>Triangle Meshes &amp; Connectivity</h4><p>Triangle meshes are generally categorized based on whether edges or triangles are considered as primary elements. An <em>Edge-based representation</em> encodes the edges of the mesh rather than the triangles themselves. It stores and manipulates edge connectivity information, making it well-suited for mesh editing operations like smoothing, refinement, and remeshing. Commonly used data structures are the <em>half-edge</em> <em>or doubly connected edge list (DCEL)</em> that stores information about each half of an edge, linking them to form a complete edge (Mäntylä, 1988) and the <em>winged- edge</em> in which each edge refers to the adjacent faces, allowing for quick access to neighboring triangles (Baumgart, 1975). Only manifold meshes can be represented using edge-based structures. A <em>Triangle-based</em> or <em>face-based representation</em> directly encodes the triangles of the mesh. <em>Index </em>structures are the most prominent methods for representing their mesh connectivity. Index data is a sequence of integers that correspond to the positions (indices) of vertices in the vertex data array. Each triplet of indices in the index data sequence represents a triangle <em>Tn</em>, with each index pointing to a specific vertex <em>vn</em> in the vertex data array. The vertex data array stores the actual coordinates <em>x, y, z </em>of the vertices, while the index data defines how these vertices are connected to form the triangles (Figure 1).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*geB8Cit9cZKLpBrn84uQDg.png" /><figcaption><strong>Figure 1</strong>. The indexed structure of a triangular mesh: vertex and index data arrays.</figcaption></figure><p>Indexed structures are more compact than edge-based representations leading to minimized data transfer and efficient caching of vertex transformations. Since vertices are referenced by indices, the GPU can store and reuse transformed vertex positions, reducing overhead and repetitive calculations. Commonly used data structures for face-based representation are <em>vertex-face</em> in which each vertex stores references to its adjacent faces, and <em>face-vertex</em> which represents each face by storing references to its vertices. Vertex-face structures facilitate vertex-based editing operations, such as smoothing, refinement and remeshing while face-vertex structures may be preferred for mesh processing, such as collision detection algorithms.</p><p>Surface attributes other than geometry are also, associated with the mesh. These attributes can be distinguished into <em>discrete </em>attributes and <em>scalar</em> attributes. Discrete attributes are typically associated with the faces of the mesh. The material identifier is a common discrete attribute that determines the shader function used in rendering a face of the mesh, e.g. simple look-up within a specified texture map. In simple cases, scalar attributes are associated with the vertices of the mesh. Examples are the diffuse color <em>(r g b)</em>, the normal <em>(nx ny nz)</em>, and the texture coordinates (<em>u v</em>). In case that adjacent faces have different shading functions, then scalar attributes at a corner (<em>v f</em>) of the mesh specify the shading parameters for face <em>f </em>at vertex <em>v</em>. For example, along a crease, i.e. a curve on the surface across which the normal field is not continuous, each vertex has two distinct normals, one associated with the corners on each side of the crease.</p><blockquote><em>If you would like to cite this work, please use: </em><strong><em>Boutsi, A.-M., 2025. Photogrammetric techniques and algorithms optimization for the development of interactive augmented reality systems. NTUA, Athens. </em></strong><a href="https://doi.org/10.12681/eadd/59886"><strong><em>https://doi.org/10.12681/eadd/59886</em></strong></a></blockquote><p><strong><em>References:</em></strong></p><p>Ponchio, F., &amp; Hormann, K. (2008). Interactive Rendering of Dynamic Geometry. <em>IEEE Transaction on Visualization and Computer Graphics</em>, <em>14</em>(4), 914–925.</p><p>Alliez, P., Ucelli, G., Gotsman, C., &amp; Attene, M. (2008). Recent Advances in Remeshing of Surfaces. In L. De Floriani &amp; M. Spagnuolo (Eds.), <em>Shape Analysis and Structuring</em> (pp. 53–82). Springer Berlin Heidelberg. <a href="https://doi.org/10.1007/978-3-540-33265-7_2">https://doi.org/10.1007/978-3-540-33265-7_2</a></p><p>Mäntylä, M. (1988). <em>An introduction to solid modeling</em>. Computer Science Pr.</p><p>Baumgart, B. G. (1975). A polyhedron representation for computer vision. <em>Proceedings of the May 19–22, 1975, National Computer Conference and Exposition on — AFIPS ’75</em>, 589. <a href="https://doi.org/10.1145/1499949.1500071">https://doi.org/10.1145/1499949.1500071</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e625f13cced1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Automation on 3D & XR web publishing — Top 10 software]]></title>
            <link>https://medium.com/@iboutsi/automation-on-3d-xr-web-publishing-top-10-software-d8a87acb6667?source=rss-cc0277c9a449------2</link>
            <guid isPermaLink="false">https://medium.com/p/d8a87acb6667</guid>
            <category><![CDATA[tools]]></category>
            <category><![CDATA[no-code]]></category>
            <category><![CDATA[web-development]]></category>
            <category><![CDATA[webxr]]></category>
            <category><![CDATA[3d-visualization]]></category>
            <dc:creator><![CDATA[Iro Boutsi]]></dc:creator>
            <pubDate>Sat, 20 May 2023 11:53:05 GMT</pubDate>
            <atom:updated>2023-05-20T11:53:05.960Z</atom:updated>
            <content:encoded><![CDATA[<h3>Automation on 3D &amp; XR web publishing — Top 10 software</h3><p>Multiple web-based frameworks offering multiple forms of data presentations, tools and interactions have arisen. Each one entails different code literacy skills, technological pipeline, data size and 3D formats. In this guide we present commercial and open-source solutions that do not require any programming skill as every asset, service and interaction is easily configurable through a Content Management System. Most well-known proprietary ones for management, preview and automatic upload are Sketchfab, echo3D, Vectary and p3d.in, offered at different pricing tiers. Meanwhile, there are several non-commercial but competitive software options available for 2D/3D data dissemination such as ATON, Resurrect3D, 5dMeteora, Kompakkt and Smithsonian Voyager.</p><h4>Commercial software:</h4><h3><strong>1. Sketchfab</strong></h3><p><a href="https://sketchfab.com/">Sketchfab</a> is probably, the most popular, easy-to-use solution for 3D models hosting and sharing, optimized for fast, on-demand loading and VR/AR inspection. It provides a cloud-based platform where users can upload their 3D models and store them securely, additional context and information via annotations as well as compatibility with VR devices or AR-enabled smartphones. It has a vibrant community of 3D artists, researchers, designers, and enthusiasts. Users can connect with like-minded individuals, engage in discussions and showcase their contributions within a team or group.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AQcJWz2iQFKHvCUU3BVuGA.png" /></figure><p><em>Pricing: </em>Sketchfab offers several pricing plans, including free use that allows users to upload and share up to 10 models per month, with a maximum file size of 100MB.</p><p>1. Pro: This plan costs $15 per month and allows users to upload and share up to 50 models per month, with maximum file size of 200 MB, model conversion and up to 20 annotations per model. Users can try a 14-day free trial of the Pro plan before committing to a subscription.</p><p>2. Premium: It costs $79 per month, it offers 200 uploads per month of 500 MB maximum size each and includes all the features of the Pro plan, plus additional advanced features such as AR apps, custom branding, advanced support, and team collaboration.</p><p>3. Enterprise: Customized for larger organizations, it includes all the features of the Premium plan, with additional features such as advanced analytics, integrations, and dedicated support.</p><h3>2. Echo3D</h3><p><a href="https://www.echo3d.com/??utm_term=echo3d&amp;utm_campaign=search_campaign&amp;utm_source=googleads&amp;utm_medium=ppc&amp;utm_term=echo3d&amp;utm_campaign=Search+Campaign+(Complete+Registration,+Any+Plan)&amp;utm_source=adwords&amp;utm_medium=ppc&amp;hsa_acc=8371579743&amp;hsa_cam=6451402957&amp;hsa_grp=145775378825&amp;hsa_ad=649144140129&amp;hsa_src=g&amp;hsa_tgt=kwd-355312257988&amp;hsa_kw=echo3d&amp;hsa_mt=b&amp;hsa_net=adwords&amp;hsa_ver=3&amp;gad=1&amp;gclid=Cj0KCQjwpPKiBhDvARIsACn-gzBYmuyrfYBeg3hbX3Gr1PQ3tNb4hn582gxMB_9fJWn1x00Ch934itwaAld6EALw_wcB">Echo3D</a> is a 3D asset management platform for developers that provides tools and cloud infrastructure to manage, update, and stream 3D content to real-time 3D/AR/VR and Metaverse applications and games. It offers a CMS and delivery network (CDN) to build the back-end system of the 3D model viewer and then, integrate into React.js and WebGL for web exploitation, AR SDKs and game engines. Besides 2D and 3D content customization and dissemination, it supports real-time collaboration, version control, custom branding and detailed analytics on user engagement and behavior.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Gb-pxqynKU-BTXDAohiRIQ.gif" /></figure><p><strong><em>Pricing:</em></strong><em> </em>echo3D uses a credit-based payment system. A credit corresponds to 5 API calls, 10MB of storage or 100 MB of bandwidth usage. Its free version provides users with 10 credits to create up to 5 projects and includes 1 GB of storage and and 1 GB bandwidth. It does not include the WebAR customizer, the Model editor, the Scene editor as well as the conversion and compression!</p><ol><li>Pro: This plan costs $19 per month and includes up to 50 projects with 1 admin and 5 users, 100 MB of storage and 10 GB bandwidth (100 credits).</li><li>Premium: This plan costs $99 per month and includes up to 500 projects with 3 admins and 15 users, 1 GB of storage and 50 GB bandwidth (500 credits).</li><li>Enterprise: This plan is designed for larger organizations and offers custom pricing. It includes advanced features such as custom branding, single sign-on (SSO), and dedicated support.</li></ol><h3>3. Vectary</h3><p>Similarly to echo3D, <a href="https://www.vectary.com/">Vectary</a> is a web publishing software with dashboard for the creation, editing and sharing of 3D experiences without any programming skills. It supports both 2D and 3D files uploading, pre-made shapes and objects, advanced settings for appearance tuning (lighting, materials, texture, effects) and ready-to-use interactions such as click and action-based events. Like echo3D, its collaboration tools allow multiple users to work on the same project simultaneously. It also, integrates with popular design tools like Sketch and Figma.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WimgasZarkEFyIkGurLIWw.gif" /></figure><p><em>Pricing</em>: Vectary offers a range of pricing plans to suit different needs and budgets. The free plan allows users to create and publish up to 3 public projects with basic features, such as basic shapes, materials, and textures.</p><ol><li>Hobbyist Plan: The Hobbyist plan costs $15/month or $120/year and offers more advanced features, such as unlimited public projects, advanced shape libraries, custom backgrounds, and export to PNG and SVG formats.</li><li>Freelancer Plan: The Freelancer plan costs $30/month or $240/year and includes all the features of the Hobbyist plan, as well as additional features like collaboration tools, high-resolution rendering, and export to STL, OBJ, and GLTF formats.</li><li>Business Plan: The Business plan is tailored for teams and organizations and starts at $80/month or $640/year for 3 users. It includes all the features of the Freelancer plan, as well as team management tools, custom branding, and priority support.</li></ol><h3>4. Verge3D</h3><p><a href="https://www.soft8soft.com/">Verge3D</a> is a toolkit, built on top of the Blender 3D software, for creating 3D models, animations and interactive experiences using a visual editor. Unlike the previous apps, it runs offline and has to be downloaded on local device. It includes a wide range of materials and textures, lighting options, physics simulations and events (mouse and touch events, keyboard input, and gamepad controllers.) for custom-made 3D experiences. It supports WebGL and WebXR, ensuring compatibility with a wide range of devices, including touchscreens, gamepads, and VR headsets. Once the 3D experience is created on local server, Verge3D allows users to publish it directly to the web or to popular platforms such as WordPress, Shopify, and others.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AVrHuLJUA9lwF8Zs8qxTdg.gif" /><figcaption>3D reconstruction of Mars Discovery Rover (source:https://eyes.nasa.gov/curiosity/)</figcaption></figure><p><em>Pricing</em>: Hobbyist: This plan is free and provides access to basic features for non-commercial use. It includes access to the Verge3D for Blender and Verge3D for 3ds Max tools.</p><ol><li>Freelancer: This plan is priced at $290 per year and includes access to all features of Verge3D for Blender and Verge3D for 3ds Max, as well as priority support.</li><li>Studio: This plan is priced at $990 per year and includes all the features of the Freelancer plan, as well as network rendering support and the ability to use Verge3D for Maya.</li><li>Enterprise: This plan is priced based on the specific needs of the user and includes custom features, support, and training.</li></ol><h3>5. p3d</h3><p><a href="https://p3d.in/">p3d.in</a> hosts and instantly visualizes online or integrates into webAR the products of 3D modelling software like Blender, 3DSMax, Rhino etc. Unlike Verge3D, it has an admin interface layout that speeds up the publishing procedure. Users can customize the appearance of their models with different materials, lighting, and backgrounds, overlay them on AR scenes and add annotations and hotspots to provide additional context and information.p3d.in also provides premium features such as password protection, analytics, and custom branding.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MtazbOYnaSC9Va8lnf3W5Q.gif" /></figure><p><em>Pricing:</em> The free plan allows to upload, edit and view in AR model 10 3D models. It is the most affordable framework of the list with only one paid subscription plan that costs $6 per month to upload up to 100 3D models, get priority support and branding. It has to be noted that it is the only one that offers a one-time payment at the cost of $60 with a lifetime of access to its premium featured.</p><h4>Open-source software:</h4><h3>6. ATON</h3><p><a href="https://osiris.itabc.cnr.it/aton/">ATON</a> is an open-source framework that lies upon Three.js library for creating cross-platform Cultural Heritage applications of high-fidelity. Since 3D models of various fields or expertise can be integrated, it isn’t content-specific. It comes with state-of-the-art modules and components for WebXR displays, querying, semantic annotations and real-time collaborative sessions. Other high-end features that make it stand-out are the advanced 3D rendering (<a href="https://en.wikipedia.org/wiki/Physically_based_rendering">PBR</a>), viewpoint (POV) management, spatial UIs for XR sessions, visual/immersive analytics and integration with complex multimedia content. ATON is actively deployed by research projects in archaeological <a href="https://www.mdpi.com/2571-9408/6/2/69">3D reconstructions</a>, <a href="https://disegnarecon.univaq.it/ojs/index.php/disegnarecon/article/view/608">museum collections</a> and <a href="https://polipapers.upv.es/index.php/vitruvio/article/view/14665">gamification</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*sM9FndmJV0BTRm1HtgUCgw.gif" /><figcaption>3D reconstruction of <a href="https://aton.ispc.cnr.it/s/vhlab/demo/aiano"><strong>Aiano Villa</strong></a></figcaption></figure><h3>7. Resurrect3D</h3><p><a href="https://engrxiv.org/preprint/view/1880/3845">Resurrect3D</a>, a customizable platform comprising a 3D rendering engine with a toolset for domain experts and a front-end infrastructure, has been proposed as an alternative of Sketchfab. Although it is oriented to Cultural Heritage models, it can be adopted by every industry or field of research that creates, owns or manages 3D files. Based on Three.js library, it focuses on UI extensibility of its basic scene editing and interaction tools. Resurrect3D by default supports a physically based rendering (PBR) workflow, the conversion tool converts 3D formats (e.g., OBJ, FBX, VRML) into a range of PBR maps (e.g., diffuse and displacement maps) in the JSON Object Scene Format required by Three.js.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/994/1*8XLMFsJfsOtCJ5D5QYrE2Q.gif" /><figcaption>3D reconstruction of <a href="https://medium.com/r?url=https%3A%2F%2Fwww.library.rochester.edu%2Fspaces%2Fstudio-x%2Fresources%2Fresurrect3d"><strong>mbira</strong></a><strong> (Sekuru’s stories)</strong></figcaption></figure><h3>8. 5dMeteora</h3><p><a href="http://meteora.topo.auth.gr/5dmeteora.php?lang=en">5dMeteora</a> framework has an integrated <a href="https://3dhop.net/index.php">3DHOP</a>’s 3D viewer, interactive search and retrieve tools, DataTables.js data sorting, categorization of information based on field or specialty as well as VR and AR capabilities. It consists of a mid-level Content Management System with an easy-to-use UI that automates publishing of 3DHOP scene’s and attachment of multimedia files (text, pdf, images and videos) to its “hotspots”. Since it adopts the multi-resolution format of <a href="https://vcg.isti.cnr.it/nexus/">Nexus.js library</a>, it supports the visualization of very large point-clouds and 3D models (up to 100 millions points or triangles). All the data are hosted into the local server of users since cloud storage isn’t supported. The framework has been designed for Cultural Heritage projects but, like the previous software, it suits to every type of 2D/3D dataset, enriched with multimedia and metadata.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BdeonT3LoDHGLojjozVuUA.gif" /></figure><h3>9. Kompakkt</h3><p><a href="https://kompakkt.de/home">Kompakkt</a> also adheres to the design principles of Sketchfab with the difference that it enables collaboratively annotation of 3D models with multimedia (Kompakkt.de, 2018). It allows users to share their own content and connect with others joining an online community. It protects the copyrights of the creators and ensures data integrity while users download 3D models. The created scenes can be embedded into any website via iframe. However, no AR or VR modes are currently, supported.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*d5HnI5p9BC_v7juuwOdURg.gif" /></figure><h3>10. Smithsonian Voyager</h3><p><a href="https://3d.si.edu/voyager-story-standalone">Smithsonian Voyager</a> emphasizes on digital storytelling through a set of customizable set of authoring and quality control tools. Users can create 3D experiences by adding 3D models to the scene, creating thumbnails and assigning a starting camera position for 3D navigations, adding multimedia, annotations, articles and other 3D tours. A scene created by Smithsonian Voyager can be hosted on its cloud or on any other external website. It is also, available offline for testing and debugging purposes by downloading a pre-built package. It currently doesn’t support AR or VR.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/958/1*9q1v4UgDNKjx3B_glOtFkA.gif" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d8a87acb6667" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A guide on Web AR performance testing (Example of AR.js — Three.js prototype)]]></title>
            <link>https://medium.com/@iboutsi/a-guide-on-web-ar-performance-testing-example-of-ar-js-three-js-prototype-492c135e74c0?source=rss-cc0277c9a449------2</link>
            <guid isPermaLink="false">https://medium.com/p/492c135e74c0</guid>
            <category><![CDATA[developer-tools]]></category>
            <category><![CDATA[benchmark]]></category>
            <category><![CDATA[augmented-reality]]></category>
            <category><![CDATA[performance]]></category>
            <category><![CDATA[web-development]]></category>
            <dc:creator><![CDATA[Iro Boutsi]]></dc:creator>
            <pubDate>Tue, 25 Apr 2023 21:24:19 GMT</pubDate>
            <atom:updated>2023-04-26T06:13:40.380Z</atom:updated>
            <content:encoded><![CDATA[<h3>A guide on Web AR performance testing with DevTools (Example of AR.js — Three.js)</h3><p>Achieving high performance on the web, particularly on mobile devices, is a challenging task due to the limited computing and rendering capabilities available. Web software can be easily, incorporated into the traditional levels of testing namely — Unit Testing, Integration Testing, System Testing, and Acceptance Testing. However, in case of web AR prototype testing, there are certain considerations that are worth knowing.</p><blockquote>Which factors should be monitored and are actually critical for AR? And, which are their relative metrics?</blockquote><p>In this guide, we investigate <strong>(i)</strong> GPU/CPU performance, <strong>(ii)</strong> memory consumption and, <strong>(iii) </strong>network load of an AR app accessed using the DevTools of Google Chrome. The proposed testing workflow is implemented into a marker-based AR prototype on the basis of AR.js and Three.js library. The digital overlay is the 3D reconstructed model of <a href="https://sketchfab.com/3d-models/middlebury-templering-ieee-spl-2022-ed66481b8f3f4b2d8ca991731ac3e4bb"><em>Middlebury TempleRing</em></a><em> </em>of 12 MB size and ply format. The code can be found <a href="https://ar-js-org.github.io/AR.js-Docs/marker-based/#threejs">here.</a></p><h3>Experimental Setup</h3><p>Tests for each factor have to be executed under the same real-life environment with the same device, codebase and conditions to ensure the accuracy and reliability of results. In case of non-static scenes, the same test may be repeated three or five times to get the average values and reduce the impact of any random fluctuations that might occur. When a series of tests is completed, then it is recommended to continue benchmarking under different environments, devices, or network conditions. The web AR experience has to be robust and reliable across a wide range of scenarios. Network throttling can be used to analyze performance in slower network profiles replicating the “worst case” scenario. It is important to identify any bottlenecks, bugs, or issues that might not be apparent in a fast network environment. Finally, CPU throttling simulates how the prototype performs on mobile devices. Each testing session is loaded after empty cache and hard reload with the following settings:</p><ul><li>CPU throttling: 4x slowdown (throttles the CPU so that it’s 2 times slower than usual)</li><li>Network: Slow 3G</li></ul><h3>Metrics and Benchmarking</h3><h4>Record Runtime Performance</h4><p>Performance profiling is achieved in the Performance Panel, by selecting <em>Record </em>and waiting for a couple of seconds (60 seconds in our case) before clicking<em> Stop.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/634/1*8D01gPcUABzXSlnQfmY4uQ.png" /><figcaption>Capturing performance metrics while the app runs.</figcaption></figure><p>To analyze the results, we have to know that<em> </em>there are <strong>three</strong> critical moments during the marker-based AR session: (i) pattern detection, (ii) loading of 3D mesh and, (iii) final display of 3D overlay. They can be identified by the events on the m<em>ain thread</em>, triggered by the relative requests of the JS functions on the <em>Network</em> section.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kVdt95By1NfdHFC_bZrUTQ.jpeg" /><figcaption>Overview of performance results regarding CPU charts and Frame rate charts</figcaption></figure><p>We inspect the following metrics at the <strong>beginning </strong>and <strong>during</strong> these moments:</p><ul><li><strong><em>CPU load &amp; Execution time</em></strong>: The colors in the CPU chart on the top of the results, correspond to the colors of the events in the main threads and in the Summary tab, at the bottom for the given time period. A <a href="https://web.dev/custom-metrics/#long-tasks-api">Long Task</a> is a task that runs on the main thread for more than 50 ms that increases CPU utilization. The user may perceive the experience as slow or unresponsive since the browser must wait for the task to finish before it can respond to any other interaction. The most direct way to examine execution time of the JS operations, is by opening the <em>Sources </em>section after selecting the desired task, as it can be seen below:</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/504/1*lgQ4LF97mF6uezBfVUBF1Q.png" /><figcaption>Debugging JS tasks of source code</figcaption></figure><ul><li><strong><em>Frame rate</em></strong>: Ideally, the AR experience should run at a smooth and consistent frame rate. To achieve 60 fps, each frame needs to be rendered in 1/60th of a second, which is approximately 16.6 ms. If multiple frames drop, which is highlighted with red in the chart, the overlay will appear jumpy, jittery, and out of sync with the real-world environment. The app’s tracking accuracy may be also affected, leading to errors in object placement and recognition.</li></ul><blockquote>If frame rates do not drop in proportion to the recorded CPU load, latency occurs during GPU processing and is driven by the rendering and not by the computing operations of the main thread.</blockquote><ul><li><strong><em>GPU utilization</em></strong>: The GPU chart is located under the main thread and measures how much of the device’s GPU is being used to render the AR content. High GPU utilization can indicate that the AR experience is pushing the limits of the device’s hardware and may lead to performance issues. An alternative way to compare frame rate with GPU load in real-time is by enabling the FPS meter. It shows the estimation of fps as the page runs as opposed to the GPU memory consumed and appears in the top-right corner of the viewport.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/192/1*OYI0uLZRf9DBa3oBphoqZA.png" /><figcaption>Frame rendering stats overlay</figcaption></figure><blockquote>Slow procedures may lie on memory leaks and undesirable memory allocation patterns. If GPU runs out of resources (OOM error) on loading a large digital overlay or running multiple processes at the same time, its important to identify the root cause.</blockquote><h4><strong>Memory profiling</strong></h4><p>Memory profiling is carried out in the Memory Panel, using the “Allocation instrumentation on timeline” type. It shows the memory distribution among JS objects over time and isolates memory leaks. These are usually caused by holding onto object references in static member variables. After selecting the desired instance, start recording after reloading the page and perform a sequence of actions (marker recognition) for a couple of seconds (60 seconds in our case) before clicking<em> Stop.</em></p><ul><li>The <strong>Summay</strong> tab shows objects total memory usage, which can be expanded to show instances. Shallow heap is the amount of memory that is held by the object itself for its description and immediate values. Thus, it is logical that arrays and strings have a significant shallow size. The retained size of the same object is the amount of heap memory that will be freed when the object is garbage collected. The heap sizes are indicators that third-party libraries or embedded resources are memory-intensive.</li><li>The <strong>Containment</strong> tab shows the objects grouped under several top-level entries and provides an overview of how much memory the app uses at a very low level. The low memory footprint may showcase increased performance in computationally intensive tasks like features detection and tracking (“<em>artoolkit</em>”) or in rendering tasks (“<em>THREE</em>”), as it can be seen bellow.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/537/1*2lalKFZ40i3uJTFakkuKtg.png" /><figcaption>“Containment” and “Statistics” sections of Memory Profiling tool</figcaption></figure><p><strong><em>Leaked objects &amp; Garbage collection</em></strong>: Garbage collection measures how often the device’s memory is being cleared of unused resources. High levels of garbage collection can indicate that the AR experience is creating and discarding large amounts of resources. To examine which operations create leaks and leave garbage, a comparison between multiple snapshots should be made as follows:</p><ol><li>Take a heap snapshot before performing an operation.</li><li>Perform an operation.</li><li>Perform a reverse operation (do the opposite interaction and repeat it a few times).</li><li>Take a second heap snapshot and change the view of this one to Comparison, comparing it to snapshot 1.</li></ol><h4>Network inspection</h4><p>Monitoring network load ensures that the AR experience is delivered smoothly and efficiently. High numbers of network requests can indicate that the app is downloading and streaming more data than necessary, which can lead to slow loading times, perceived latency and bottlenecks. Recording the network load in the Network panel, for the desired time period, results to a simple log with the number, the duration and the size of network requests. At the bottom of the panel, DevTools lists the total size of transferred and loaded (uncompressed) resources. In case of the AR experience, the request with the longest duration and the largest size (combined size of the response headers plus the response body) concerns the 3D mesh. Large values of the content download indicate that the browser is busy or the network is slow.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/696/1*FNnbAn1dB8EssVzFdAwe4A.png" /><figcaption>Network section with the waterfall and the log of the resources, listed chronologically.</figcaption></figure><p>Before releasing the app, its important to investigate if and how much this delay while reading the response, degrades the end-user experience and then, optimize the underlying code appropriately. The app production implies running numerous operations simultaneously and respond to different requests from a variable number of users. Executive load and stress testing for prototypes (on localhost) can be configured and implemented with tools like k6. They assess its current performance under typical and peak load as well as, its maximum capacity in terms of users or throughput.</p><p><strong><em>Note:</em></strong> <em>Taking into account the fact that a marker-based AR application should act in real-time, the image matching process between the pattern image and consecutive frames should be accomplished in the shortest time possible. Hence, some important measurements regarding registration and tracking, can be:</em></p><p><em>(i) the computational time of the image matching process, which is the most resource-demanding task of the computer vision workflow,</em></p><p><em>(i) the average time required by the four basic steps of the image matching and outlier removal process (i.e., feature detection; feature description; main process of image matching; and homography estimation for geometric verification of matches),</em></p><p><em>(ii) the number of feature points detected in the last frame of each sequence,</em></p><p><em>(iii) the registration accuracy, analyzed using the root mean square error (RMSE), of the distances between the inliers and their reprojected 3D reference features using the refined estimation of the camera pose.</em></p><p><em>The impact of additional parameters like lighting conditions, camera distortion, occlusion and noise may be taking into account. Tests should be executed with high scene diversity and by taking into account various sensor inputs like inertial, radio or depth data. For example, a possible evaluation needs to be undertaken for the IMU tracking errors accumulated over time in relevance to translation, rotation and distance from the camera. Finally, cross-platform applications may require both head-mounted and hand-held AR devices benchmarking.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=492c135e74c0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What’s next for Augmented Reality? Issues & challenges]]></title>
            <link>https://medium.com/@iboutsi/whats-next-for-augmented-reality-issues-challenges-c0d515304af2?source=rss-cc0277c9a449------2</link>
            <guid isPermaLink="false">https://medium.com/p/c0d515304af2</guid>
            <category><![CDATA[immersive-technology]]></category>
            <category><![CDATA[augmented-reality]]></category>
            <category><![CDATA[webxr]]></category>
            <category><![CDATA[3d-rendering]]></category>
            <category><![CDATA[extended-reality]]></category>
            <dc:creator><![CDATA[Iro Boutsi]]></dc:creator>
            <pubDate>Fri, 24 Mar 2023 07:26:55 GMT</pubDate>
            <atom:updated>2023-04-25T21:27:28.732Z</atom:updated>
            <content:encoded><![CDATA[<p>Augmented Reality (AR) allows for displaying virtual information aligned with their real-world position through a camera-equipped device. At its core, AR is a motion tracking issue involving the concepts of features detection and matching, photogrammetric bundle adjustment, homography estimation and 6 Degrees of Freedom (DoF) tracking. It entails stable poses and accurate localization across successive image frames derived from the camera feed in real-time. While it keeps pace with high-end mobile devices and their powerful sensors, there are still certain challenges need to be addressed for the ultimate immersive experience.</p><h4><strong>Occlusion:</strong></h4><p>The goal of AR is the seamless integration of a real scene with additional digital data in order to impact the sense of what is real. Occlusion refers to the effect of real-world objects blocking the view of virtual objects. Real-time occlusion handling involves quick and accurate acquisition of depth maps of the real scene during the AR session. At present, depth estimation can be achieved through depth camera, deep learning and monocular images. Besides the variety of solutions provided, it is a long-standing problem, especially for marker-less outdoors AR cases.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/679/1*uCVw2Uo8uvCNTnxXeGk1mQ.png" /><figcaption>Effective occlusion handling leveraging the LiDAR scanner of iPhone 14 Pro</figcaption></figure><h4>Tracking technologies:</h4><p>Computer vision methods based solely on features points like Structure from Motion (SfM), often lead to large registration errors while localization with visual Simultaneous Localization and Mapping (vSLAM) or Visual Odometry (VO) is prone to partial occlusion. On the other hand, outdoors GPS positioning is still inaccurate and subject to shifting by a certain offset and coordinate changes over time. Although Deep Learning (DL) underlies end-to-end solutions to 6 DoF pose estimation, temporal stability of the tracking is hard to handle by the neural network directly.</p><h4><strong>Streaming &amp; 3D rendering:</strong></h4><p>Typically, the overlaid multimedia information comprises text, animation, 3D graphics as well as CAD, wireframe and low-poly models. However, the current advances in photogrammetry, range sensors and fast surface modelling produce high-resolution, geometrically accurate and realistic 3D reconstructions. Their large size and complex structure dictate their integration into AR applications, where visual fidelity is encountered as a crucial factor. The obvious trade-off between visual quality and overall responsiveness is an issue that needs to be tackled even by the commercial AR SDKs. Rendering can be offload to cloud servers minimizing overhead but data transfer suffers from E2E latency and the risk of security and<br>privacy threats. MEC-oriented web AR solutions compensate for this latency as well as the weak computational efficiency of browsers through the operation of a client-server model typically located close to the user’s device. However, both technologies are prone to single point failure and<br>network load while huge or multiple chunks of 3D data are transferred.</p><h4>Lack of standards:</h4><p>While AR is still relatively new and rapidly evolving, there is no single standard that is universally accepted or adopted. One of the most widely used standards is the Open Geospatial Consortium’s (OGC) <a href="https://www.ogc.org/standard/arml/">Augmented Reality Markup Language</a> (ARML), which is an XML-based language for describing AR content and interaction. ARML is designed to enable interoperability between AR systems and applications, allowing developers to create AR content that can be shared and displayed across a variety of platforms and devices. Another popular standard is the <a href="https://labs.mozilla.org/projects/webxr-viewer/">WeXR API</a>, supported by a growing number of browsers, including Google Chrome, Firefox, and Safari. It builds cross-platform AR experiences that can be accessed through web browsers without the need for a dedicated app or plugin. Other AR standards include the <a href="https://learn.microsoft.com/en-us/windows/mixed-reality/develop/native/openxr">OpenXR</a> standard from <a href="https://www.khronos.org/">Khronos</a> Group for creating immersive and interactive experiences across a range of devices, and the <a href="https://mpeg.chiariglione.org/standards/mpeg-a/augmented-reality-reference-model/wd-isoiec-23000-14-augmented-reality-reference">ISO/IEC 23000–14</a> standard for multimedia application formats. Another issue is the lack of a lightweight 3D format file suited to AR overlays. The emergence of dedicated mobile web AR browsers, such as <a href="https://gvu.gatech.edu/research/projects/argon-ar-enabled-web-browser">Argon</a> and WebXR, has not resolve the compatibility issues that derive from the absence of a standard for Web 3D objects.</p><h4>Hardware requirements:</h4><p>High-end smartphones and tablets have turned into a necessity both for mobile AR and web AR apps. They incorporate depth APIs for depth maps using stereo cameras, structured light, 3D Time of Flight (ToF) depth sensors, scanner-less LiDARs, multiple GPU texture resolutions and 5G networks. Sensors have their strengths and weaknesses, and their effectiveness depends on the specific application and environmental conditions in which they are used. ToF are IR (infrared) based sensors that are prone to noise and inaccurate measurements during bright sunlight. Stereo cameras are not affected by the same interference from sunlight that IR sensors are, but they do require a well-lit environment with lots of features and contrast to function properly. The most popular AR SDKs, like ARCore, ARKit and Vuforia, are running on device models that fulfil certain minimum hardware requirements, such as quad-core CPUs and the existence of Inertial Sensor Units (IMU) sensors. The quality of the camera, the version of the operating system and the architecture also determine the compatibility.</p><p><strong><em>References:</em></strong></p><p>M. Macedo and A. Apolinario, “Occlusion Handling in Augmented Reality: Past, Present and Future,” <em>IEEE Transactions on Visualization &amp; Computer Graphics</em>, vol. 29, no. 02, pp. 1590–1609, 2023. doi: 10.1109/TVCG.2021.3117866</p><p>S. Verykokou, A.-M. Boutsi, and C. Ioannidis, “Mobile Augmented Reality for Low-End Devices Based on Planar Surface Recognition and Optimized Vertex Data Rendering,” <em>Applied Sciences</em>, vol. 11, no. 18, p. 8750, 2021. doi: 10.3390/app11188750.</p><p><a href="https://medium.com/hackernoon/why-is-occlusion-in-augmented-reality-so-hard-7bc8041607f9">https://medium.com/hackernoon/why-is-occlusion-in-augmented-reality-so-hard-7bc8041607f9</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c0d515304af2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Sustainable planning for Cultural Heritage in 3D digital era — Tools & Opportunities]]></title>
            <link>https://medium.com/@iboutsi/sustainable-planning-for-cultural-heritage-in-digital-era-e9e717003fb5?source=rss-cc0277c9a449------2</link>
            <guid isPermaLink="false">https://medium.com/p/e9e717003fb5</guid>
            <category><![CDATA[visualization]]></category>
            <category><![CDATA[sustainable-development]]></category>
            <category><![CDATA[3d-modeling]]></category>
            <category><![CDATA[cultural-heritage]]></category>
            <category><![CDATA[web-development]]></category>
            <dc:creator><![CDATA[Iro Boutsi]]></dc:creator>
            <pubDate>Mon, 20 Mar 2023 12:14:44 GMT</pubDate>
            <atom:updated>2023-03-20T12:15:41.237Z</atom:updated>
            <content:encoded><![CDATA[<p>Cultural Heritage (CH) sustainability has been recently introduced to <a href="https://www.cepal.org/sites/default/files/events/files/2030_agenda_and_the_sdgs_an_opportunity_for_latin_america_and_the_caribbean.pdf">Agenda 2030</a> to describe the relationship between conservation and sustainable development. Goal 11 advocates “Strengthen efforts to protect and safeguard the world’s cultural and natural heritage”. Conservation plays a significant role in sustaining local communities; not only reinforcing local identity and traditions, but also, maximizing economic benefits through well-managed tourism. Besides the social and economic dimension, CH sustainability tackles the harmful effects of urbanization and tourism on the environment. Tourism is a conflicting factor that can easily override the natural carrying capacity of a site. It may lead to overcrowding, traffic congestion, ecological deterioration and local resident’s dissatisfaction. Planning is essential to determine the limits of natural and cultural environment, as opposed to the maximum number of visitors. While problems related to carrying capacity have to be diagnosed, the conservation of CH authenticity must counterbalance priorities on economic and market values.</p><p>To deal with these threats and recognize the opportunities, sustainable assessment and management have to be conducted. Planning, decision making, regulatory and standards development as well as knowledge transfer are mandatory but inefficient if they are not accessible, visually perceived and participatory. In this perspective, photogrammetry has a crucial role in documenting, digitally preserving and promoting CH assets through the generation of 3D models. The documentation of cultural diversity against the trend of globalization is important to gain insight into any of these policies or actions of conservation. Not only it reveals their universal significance and originality, but also it enables an immediate and intuitive representation of current state. Once published on the web, the 3D models bridge the gap between stakeholders involved in CH management. They can be used as a cartographic basis to link geopolitical, economic and social operations with their spatial reference. The most common and long-established technologies for spatial analysis are the following ones: (i) <strong>Geographic Information System (GIS)</strong>; (ii) <strong>Geocoding</strong>; and (iii) <strong>Spatial Database Management Systems (SDBMS)</strong>. To offer more than simple 2D mapping and geocoding, a collection of software that orients to WebGL API’s 3D rendering, dynamic uploading and organization of multi-modal datasets as well as interactive data retrieval is presented here:</p><h4><em>No-code development — GIS based CH:</em></h4><h3>ArcGIS Web AppBuilder</h3><p><a href="https://developers.arcgis.com/web-appbuilder/">ArcGIS Web AppBuilder</a> is part of the ArcGIS developer tools of ESRI’s ArcGIS that allows to build a fully featured HTML app. Besides the uploading and online dissemination of local GIS geodatabases, it creates new map layouts and integrates of multiple non-GIS data types, including 3D models, interactive services and collaboration options. For CH data management, it offers powerful functionalities for spatial analysis and representation modes of semantics, metadata and multimedia in association with their spatial dimention:</p><ol><li><em>Geoprocessing tools: </em>Widgets like query, suitability modeler, measurement, proximity analysis, data transformation etc. offer real-time manipulation and analysis of the geodata. For example, the end-user can create buffers around historic sites, calculate distances between landmarks, and generate reports on the spatial relationships between different CH features.</li><li><em>Feature editing</em>: With feature editing tools, the end-user can easily update and maintain the CH data for collaboration, participation in decision-making and/or crowdsourcing purposes. This includes adding new sites or features, updating existing data, and deleting obsolete information.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TMyYX5FwdnIPucWvwPLvdw.png" /><figcaption>Environment of ArcGIS Web AppBuilder with ready-to-use widgets, themes and interactive services for dynamic preview and automatic publishing</figcaption></figure><p>Regarding 3D visualization, 3D models can be imported into feature classes of map packages and displayed into the scene. It is recommended to use Khronos Group’s Graphics Language Transmission Format (.gltf or .glb) which is gaining adoption as the standard for transmitting 3D models over the Web. However, low-poly, CAD and BIM models are currently supported, mainly derived from Blender, Autodesk Maya, and Trimble SketchUp.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/997/1*L4zxGel9QDw1FtnFC5Usxw.png" /><figcaption><a href="https://www.esriuk.com/en-gb/arcgis/3d-gis/overview">3D city visualization</a> for San Francisco</figcaption></figure><p><em>Pricing</em>: Pricing is scalable based on the deployment option and licensing model chosen or the number of location services needed. If you are a user of ArcGIS Online, you don’t need to pay any additional fees beyond your subscription. Research or academic institutions may qualify for free access to ArcGIS software. Esri also offers a Developer Edition of the Web AppBuilder that can be downloaded and used for free for non-production purposes. This version is intended for internal use or for deployment to a limited audience.</p><p><em>Extra tips</em>: For customizable widgets and themes, a version of ArcGIS Web AppBuilder for developers is available. For achieving even more extensibility, <a href="https://developers.arcgis.com/javascript/latest/">ArcGIS API for JavaScript</a> or for <a href="https://developers.arcgis.com/python/">Python</a> may suit to more experienced developers while third-party libraries like <a href="https://developers.arcgis.com/esri-leaflet/&#39;">Leaflet.js</a> and <a href="https://developers.arcgis.com/maplibre-gl-js/">MapLibre GL JS</a> offer APIs for accessing ArcGIS location services.</p><h3>MapBox Studio</h3><p>Mapbox Studio is the web version of the Mapbox Studio platform with user-friendly GUI. It is a cloud-based map design tool that allows to build, style, and publish interactive maps, map data, and location-based services. Besides the creation of custom maps that highlight cultural heritage sites, such as museums, archaeological sites, and historic landmarks, it provides a useful set of tools for management and planning. However, in terms of geoprocessing features, it is quite limited in comparison with ArcGIS Web AppBuilder.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*Z7Ly9wd9OYeYNasJmG6eDw.gif" /><figcaption>Customization of vector data in the Mapbox Studio style editor</figcaption></figure><ol><li><em>Labels, annotations &amp; POIs</em>: They are clickable chartographic symbols, placed on top of the basemap that identify quantitative and qualitative properties, as well as links to related information and multimedia. They mark the location of historic landmarks, museums, or other cultural heritage sites, promote lesser-known monuments or highlight specific landmarks making it easier for visitors to find and explore the CH sites.</li><li><em>Data conditions/filters</em>: The filter property of a layer specify conditions that determine which features are displayed while Filter Editor creates more complex conditions based on attribute values, location, and more. In addition to filtering, Mapbox also supports querying features based on end-user input or other events.</li><li><em>Feature editing</em>: The end-user can add POIs, cartographic symbols and route lines to provide additional context about specific heritage sites or artifacts. Thus, they can be involved in monitoring ad-hoc by reporting the current state and supplement CH maps with walking and hiking routes, exact location data and information regarding local touristic businesses, descriptions and statistics as well as general information about touristic impact and level of heritage attractiveness.</li><li><em>Elevation profile</em>: It displays elevation changes of a selected path or line, helping end-users to visualize its steepness, and any hills, valleys or other changes in the terrain. It also highlights specific points along a trail such as trailheads and analyzes its characteristics such as the maximum and minimum elevation, length etc. It is particular useful in spatial analysis, route planning as well as for the evaluation of the suitability of a trail and the identification of potential safety concerns for land managers.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/873/1*1sINYHmzyiStKBlfLeEN5g.png" /><figcaption>CH POIs visualization &amp; information retreival in Sydney with Mapbox Studio</figcaption></figure><p><em>Extra tips</em>: The display of 3D models on top of the map layer entails a bit of coding. Developers can exploit <a href="https://docs.mapbox.com/mapbox-gl-js/guides/">Mapbox GL JS</a> that uses Three.js library and WebGL to render 3D content and interactive maps from vector tiles. A quick tutorial can be found <a href="https://docs.mapbox.com/mapbox-gl-js/example/add-3d-model/&#39;">here</a>. Unlike the requirement of some programming, the integration of Three.js is giving the flexibility of handling 3D geometries of various sizes, formats and complexity.</p><p><em>Pricing</em>: Maps SDK of MapBox is free for the first 50K monthly requests and then, the pricing model is <a href="https://www.mapbox.com/pricing/#maploads">pay-as-you-go</a>.</p><h4>No-code development — 3D CH repository:</h4><h3>Sketchfab</h3><p>Sketchfab is a popular, easy-to-use solution for 3D models hosting and sharing, optimized for fast, on-demand loading and VR/AR inspection. It comprises the category “<a href="https://sketchfab.com/store/3d-models/cultural-heritage-history?ref=header">Cultural Heritage &amp; History</a>”, explicitly designed to be deployed for the CH sector. It supports a variety of file formats such as <a href="https://sketchfab.com/store/3d-models/cultural-heritage-history?file_formats=stl">STL</a>, <a href="https://sketchfab.com/store/3d-models/cultural-heritage-history?file_formats=ply">PLY</a>, <a href="https://sketchfab.com/store/3d-models/cultural-heritage-history?file_formats=fbx">FBX</a>, <a href="https://sketchfab.com/store/3d-models/cultural-heritage-history?file_formats=obj">OBJ</a>, <a href="https://sketchfab.com/store/3d-models/cultural-heritage-history?file_formats=3ds">3DS</a>, <a href="https://sketchfab.com/store/3d-models/cultural-heritage-history?file_formats=max">MAX</a>, <a href="https://sketchfab.com/store/3d-models/cultural-heritage-history?file_formats=c4d">C4D</a>, <a href="https://sketchfab.com/store/3d-models/cultural-heritage-history?file_formats=blend">BLEND</a> and GLTF and offers the following functionalities regarding CH planning:</p><ol><li><em>Annotations</em>: They highlight specific details or features of an artifact, provide historical or cultural background information, or explain the significance of a particular building or site. They facilitate collaboration among domain experts in order to share knowledge and connect policies with their spatial reference. They can also be used to create interactive tours or presentations, allowing simple users to explore a CH site in a more immersive and engaging way as well as build a more comprehensive understanding of measures, policies and strategies.</li><li><em>Animations</em>: An animated navigation creates a fly-through of a CH site or building and illustrates different aspects of a historical event or cultural tradition. They illustrate the current function or significance of different features of a CH site, such as the layout of a city or the construction of a temple. In case of decision-making, they can be used to showcase how a particular artifact was prior to restoration, how a building may have looked in different historical periods or how the implementation of certain policies will evolve a CHsite in the future.</li><li><em>Analytics</em>: Sketchfab provides analytics tools that can help to track viewer engagement and assess the impact of the uploaded CH 3D models. This information can be useful for making informed decisions about how to improve and promote these resources.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1004/1*Fzrc2zzrSo16xYjgGp1C4Q.png" /><figcaption>Annotations with short description on the <a href="https://sketchfab.com/3d-models/reconstruction-of-the-monastery-of-st-modestos-754ce9b619d84e478cacb1cec25df28b">3D model</a> of Meteora Project</figcaption></figure><p><em>Pricing</em>: It has three pricing plans, from free to 79 $ per month billed yearly. The basic plan includes 10 uploads per month, 100MB maximum file size and 10 annotations per model. However, Sketchfab gives the potential for small institutions with minimal resources to disseminate their own 3D digitised content. In 2020, it launched the program “Sketchfab for Museums and Cultural Heritage” authoring business accounts for museum professionals and cultural institutions including the Minneapolis Institute of Art, Chile’s Museo Nacional de Historia Natural and the University of Dundee Museum Collections in UK.</p><h3>Vectary</h3><p><a href="https://www.vectary.com/">Vectary</a> is a web publishing software with dashboard for the creation, editing and sharing of 3D experiences without any programming skills . It supports both 2D and 3D files (OBJ, STL, FBX, GLTF, GLB, DAE, IGS/IGES, STP/STEP formats) uploading and advanced settings for geometry convertion and appearance tuning (lighting, materials, texture, effects) on the basis of Three.js library. Unlike Sketchfab, it doesn’t have an asset category oriented to CH sector. However, for CH management and dissemination of multiple types of sustainable measures it offers the following tools:</p><ol><li><em>Interactions &amp; Animations: </em>Interactions comprise ready-to-use animation action controls, visibility changes, click/hover events and conditions that can create a complete story-telling. This type of dynamic and engaging presentation helps to better understand the structure and context of CH sites, making it easier to identify potential issues or areas for improvement.</li><li><em>Floating UIs</em>: The integration of relative multimedia, links, metadata, statistics etc. in the 3D scene enables stakeholders to make more informed decisions about proposed changes to CH structures. They may report the current state and illustrate the potential impacts of proposed policies and decisions on sustainable development. Especially if a comprehensive presentation of the products of a geometric documentation is intended, the heterogeneous dataset can be exposed in its entirety through the floating UIs.</li><li><em>Hotspots</em>: They are 2D elements that can be placed on the surface of the 3D model. Once clicked or hovered over, they activate various events defined as Interactions and Animations. Apart from enhancing and enriching the visualized CH assets, they communicate the spatial interconnections to the end-user.</li><li><em>Collaboration &amp; adding comments</em>: It is a valuable feedback feature that simplifies communication, collaboration and participation in decision-making process. It allows multiple users to work on a project simultaneously, especially on large-scale CH assets that require input from multiple stakeholders.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hqO2ZD0FDBvabs1ZM6OuEQ.png" /><figcaption>Environment of Vectary studio with scene customization and 3D model visualization settings</figcaption></figure><p><em>Pricing</em>: It has two pricing plans; the free use limited to 10 projects and Pro version for unlimited projects that costs 15$ per month billed annualy. The features mentioned above (animations, interactions, AR and version history) are only available in the Pro version.</p><h3>Resurrect3D</h3><p><a href="https://github.com/rochester-rcl/resurrect3d">Resurrect3D</a> is a browser-based 3D viewer currently in development by the Digital Scholarship Lab and the Lazarus Project. It is designed specifically as an open source alternative to services such as Sketchfab, with a suite of tools designed to facilitate scholarship and research of 3D CH models. It stores all of its 3D data in Three.js <a href="https://github.com/mrdoob/three.js/wiki/JSON-Model-format-3">JSON Model Format</a> and embeds a geometry conversion tool like Vectary.</p><ol><li><em>Annotations</em>: A simple symbol on top of the 3D models indicates that specific points of interest come with a short description. The list of annotations appears in the shortcuts section of the related toolbar. This multi-level correlation of 3D and 2D cultural data elucidates technical assumptions and meanings, outlines the research findings and depicts the outcomes.</li><li><em>Dynamic presentation</em>: Annotations can also synthesize an exploratory walk-through large-scale visualizations and each session can be saved in order to be replayed by end-users. The current camera position and light settings at the time of saving will be associated with the given annotation. The end-users have the opportunity to explore all the points or areas of interest of a CH site in short time with descriptive captions.</li><li><em>Measurement suit</em>: It allows end-users to take interactive measurements of the surface of the 3D model in the original units that it has been georeferenced or scaled to.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/985/1*mNBActommPOmj5wRU-xzrQ.png" /><figcaption>The 3D reconstruction of Sekuru Chigamba’s primary instrument (<a href="https://www.library.rochester.edu/spaces/studio-x/resources/resurrect3d"><em>mbira dzavadzimu</em></a>)</figcaption></figure><p><em>Extra tips</em>: Unlike its commercial counterparts, Resurrent3D exposes a set of programming interfaces to developers to extend the toolbox.</p><h4>Basic programming skills required— 3D high-res visualizations:</h4><h3>Potree</h3><p><a href="https://github.com/potree/potree">Potree</a> is an open-source JavaScript library for visualizing large point clouds on the web. It was developed by Markus Schütz as a research project at the Institute of Computer Graphics and Algorithms at the Vienna University of Technology. Point clouds are typically captured using 3D scanning technologies such as TLS, lidar or structured light or derive from image-based photogrammetry. It has since become widely used in the fields of geospatial and environmental data visualization, as well as in industries such as architecture, engineering, and construction. Potree is completely extensible and customizable, with a range of plugins and options for developers to modify and enhance its functionality. For spatial analysis and dissemination of sustainable policies, it supports the following tools:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mMEcaZ7JP7KemkqkxXALsg.png" /><figcaption>Point cloud of the <a href="http://potree.org/potree/examples/showcase/retz.html">Town of Retz</a> in the context of Potree’s viewer</figcaption></figure><ol><li><em>Geospatial integration</em>: Potree integrates geospatial data in their correct geographic location, such as maps, vectors or aerial imagery, allowing for visualization and analysis of CH point clouds and spatial data of multiple sources in the same coordinate system. It supports shapefiles, geopackages and files of common GIS formats. Thus, planning efforts take into account factors such as environmental impact, land use, and accessibility.</li><li><em>Selection and cross-sectional analysis</em>: End-users can select individual points of the visualized CH asset in order to be highlighted, labeled, or used to perform measurements or other analysis functions. By clustering data points, information about the CH site’s condition, conservation status, and management needs can be added. This is extremely useful for planning conservation activities and allocating resources to different areas of the site. Cross-sectional analysis of point clouds is also supported, enabling planners and managers to identify structural issues, such as cracks or bulges, that may not be visible in traditional 2D images.</li><li><em>Navigation and animations</em>: It supports various navigation modes (first-person, free flight, orbit, orhographic etc.) and fly-through animations taking viewers on a guided tour of the CH site or asset. Both stakeholders and non-experts gain a better understanding of the site’s layout, topography, and spatial relationships between different features. More advanced are the time-lapse animations that show changes in the data over time and animated cross-sections of the previous tool, that reveal the internal structure of the data for site assessment and planning. The animation functions can also be used to monitor and evaluate the effectiveness of conservation and management measures over time.</li><li><em>Measurements and elevation profile</em>: Besides distances, the end-user can measure areas, angles and volumes directly in the point cloud. The elevation profile is a novel feature for geospatial analysis, land use planning and environmental monitoring. It allows to create a graph of the elevation data along a selected path, like the one supported by ArcGIS Web AppBuilder. Areas that may require conservation or protection measures can be accurately identified while the effects of climate change can be monitored.</li><li><em>Annotations</em>: Annotations, placed at specific locations in the point cloud can be viewed and edited by end-users at the same time. Text, images, links and video URLs can be displayed and/or added to individual points, providing context about historical or cultural significance, conservation status, or management needs. The annotation panel offers filtering and searching services for specific annotations based on end-users content or location.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Yd__L7yxgzuGB5tnSYHvqw.png" /><figcaption><a href="http://For example, changes in elevation could indicate erosion or other forms of damage to a site.">Elevation profile tool</a> of Potree: graph of the height profile of the drawn section</figcaption></figure><h3><strong>3DHOP framework</strong></h3><p><a href="https://3dhop.net/index.php">3DHOP</a> is an open-source framework, developed by <a href="http://vcg.isti.cnr.it/">Visual Computing Laboratory</a> — <a href="http://www.isti.cnr.it/">ISTI</a> — <a href="http://www.cnr.it/">CNR</a> that is dedicated to the visualization of CH 3D models and point clouds. By using a multi-resolution 3D format, it is able to work with high-resolution 3D models (hundreds millions of triangles/points) with ease, also on low-bandwidth. The 3D models can be used as the reference surfaces that illustrate the current state or situation in high detail and metric accuracy. Built-in tools that illustrate specific management needs and promote sustainable policies are the following:</p><ol><li><em>Hotspots</em>: They are clickable 3D models or point clouds that are rendered with a solid color with alpha channel. They can be placed anywhere in the basic 3D map indicating areas of the site that require special attention or that are particularly vulnerable to damage or degradation. Once the end-user selects a hotspot, related multimedia and metadata are shown in a floating UI. This information can help stakeholders to better understand the site and its significance, to identify patterns or trends that may not be immediately apparent, to build consensus and facilitate more effective decision-making.</li><li><em>Measurement &amp; sectioning suit</em>: Like Resurrect3D and Potree, it enables the extraction of the original coordinates of any selected point on the model’s surface and the Euclidean distance measurement between two selected points. A novel feature is the cross-section toolset for real-time orthogonal sections through the 3D scene or specific models. This can be especially useful for geospatial scientists to identify hidden or hard-to-see features that may be difficult to detect using other methods.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/785/1*hhapDFegWFSGwHxfKiG1VA.png" /><figcaption>3D web-embedded viewer of 3DHOP for high-resolution models and point-clouds</figcaption></figure><p><em>Extra tips</em>: 3DHOP is efficient for streaming and visualizing online photogrammetric 3D meshes, characterized by their geometric complexity, huge size, metric precision and detailed texture mapping. However, just like Potree, it lacks a management system for dynamically setting up and updating 3D scene entities as well as authoring services. A recent work that automates publishing through a content management system, without the need of coding, has been proposed by <a href="https://meteora.topo.auth.gr/5dmeteora.php?lang=gr">Meteora project</a>.</p><p>The 3D representations are common ground among the involved parts (professionals, researchers, institutions, organizations) for integrating and synthesizing various different approaches, methodologies and practices. Policymakers can make more informed decisions about conservation and development that balance the need for economic growth with the need to protect CH. The aforementioned software enable comprehensible planning, managing, monitoring and decision-making process related to CH sites, cooperation with internal and external partners &amp; involvement of citizens and stakeholders in relevant decisions, interactive visualization of 2D/3D spatial models and open-access possibility for various types of data (spatial/non-spatial). All can be exploited as stand-alone software or be embedded into any HTML website to finally, engage the public in discussions about heritage sites and sustainable policies.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e9e717003fb5" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>