<?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 Loni Olowookere on Medium]]></title>
        <description><![CDATA[Stories by Loni Olowookere on Medium]]></description>
        <link>https://medium.com/@Loni.O?source=rss-6fed2e13225a------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*AP6P4E5K1jE9IVp6-rWrtQ.jpeg</url>
            <title>Stories by Loni Olowookere on Medium</title>
            <link>https://medium.com/@Loni.O?source=rss-6fed2e13225a------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 18:46:19 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@Loni.O/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[Types of Data Annotation: Bounding Boxes]]></title>
            <link>https://medium.com/@Loni.O/types-of-data-annotation-bounding-boxes-689a676ea2c?source=rss-6fed2e13225a------2</link>
            <guid isPermaLink="false">https://medium.com/p/689a676ea2c</guid>
            <dc:creator><![CDATA[Loni Olowookere]]></dc:creator>
            <pubDate>Fri, 08 Apr 2022 13:52:30 GMT</pubDate>
            <atom:updated>2022-04-08T13:52:30.331Z</atom:updated>
            <content:encoded><![CDATA[<p>Building an artificial intelligence (AI) model that is accurate and free from bias requires several crucial processes. One of such crucial processes is known as data annotation. <a href="https://datagen.tech/blog/data-annotation/">Data annotation</a> involves the labeling and categorization of large volumes of data to train AI models for specific use cases. There are many types of techniques used in data annotation depending on the format of the data involved.</p><p>In this article, I take a brief dive into one of the most prominent annotation techniques used in computer vision AI models known as bounding boxes.</p><h3>Bounding Boxes</h3><p>Bounding boxes is the most basic type of data annotation. It is commonly adopted because of its ease and flexibility in practice. This annotation technique works by drawing a rectangle or square around the target object, using the x and y-axis coordinates in both the upper-left and lower-right corners of the rectangle to detect objects and locations. This is particularly useful when the objects are relatively symmetrical or when the exact shape of the object is of less interest.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/802/0*aLECN1qEJikketl-" /></figure><p>Source: Google Images.</p><h3>Applications and Use Cases</h3><p>Bounding boxes are widely used in computer vision AI models. It is much needed in object classification of images and to localize images for computer vision. It can also be used for multiple use cases requiring semantic segmentation like retail clothing, autonomous vehicle driving, satellite images and furniture detection etc. Let’s explore some industrial applications of bounding boxes:</p><ol><li><strong>Autonomous cars</strong></li></ol><p>Bounding boxes are used in the training of self-driving cars to identify the different types of objects that you can see on the road like traffic signals, sidewalks or lane obstacles. These visible objects can be annotated with bounding boxes to make them easily recognizable for machines to comprehend the surroundings and prevent vehicular crashes on roads.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/252/0*FUD2R6dbPG56eGn0" /></figure><p>Source: Google Images</p><p><strong>2. E-commerce</strong></p><p>Bounding boxes are also used for products that are sold on online retail platforms. The bounding boxes can be used to recognize clothing items and other fashion accessories bought by a customer. Any online retail item can be quickly annotated with the bounding box technique to assist visual search machine learning models to identify such items and offer the required information to customers.</p><p><strong>3. Drone and robotics imagery</strong></p><p>Images can be annotated with the help of bounding boxes to label objects from the viewpoint of robots and drones. These images help robots and drones to identify the wide variety of objects seen on earth either for surveillance or geospatial purposes. You can find out more about this through this <a href="https://humansintheloop.org/daedalean-case-study/">use case</a>.</p><h3>Bounding Boxes’ Best Practices</h3><p>As simple as drawing a box around an object might seem, it still has to be done correctly to achieve desired results. Sub-optimal data annotation and inconsistencies can lead to the underperformance of computer vision AI models. The following tips below are important practices to take note of when annotating with the bounding boxes technique.</p><ul><li><strong>Use tight bounding boxes. </strong>Bounding boxes should be tight around labelled objects. This is to help the AI model learn the precise coordinates of images.</li><li><strong>Avoid diagonal items. </strong>Boxing diagonal items is a bit problematic because they take up significantly smaller space within the bounding box compared with the background. This is why diagonal items are best labelled using polygons and instance segmentation instead.</li><li><strong>Ensure pixel-perfect tightness. </strong>The edges of bounding boxes must cover the outmost pixels of the image that is being labelled. Not ensuring tightness leaves gaps that creates several IoU discrepancies Intersection over Union also known as IOU can be described as the area of overlap between your model’s prediction and the ground truth, divided by their union.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/0*qzVTZ7N_YeQk-2pZ" /></figure><p>Source: Google Images</p><ul><li><strong>Maintain variety in box-sizing. </strong>Do not forget that your model will not perform well if it is only trained on objects of the same size. Experimenting with many variations of the same object ensures that the model’s detection abilities remain optimal.</li><li><strong>Reduce box overlap. </strong>It is important to avoid annotation overlap at all times. This is because objects that have a labeling overlap with other items end up as clutter. The model will fail to distinguish the target element effectively if overlaps with another entity are excessive.</li></ul><h3>Limitations</h3><p>While bounding boxes are the first thing that comes to mind when thinking of image annotation, it is not always the best option in all cases. For example, we have slightly touched on the fact that bounding boxes are relatively inadequate in the labeling of diagonal items compared to polygonal segmentation. Another limitation is that 2D and 3D bounding boxes sometimes include background pixels that are not part of an object. This can distort model training in many ways. Finally, when using the bounding boxes technique, annotators are likely to assume or exaggerate the spatial depth of images. This can cause setbacks to model training as well.</p><p>In conclusion, bounding boxes is a very common technique in the annotation of computer vision AI models.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=689a676ea2c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Noya: Unlocking the Hidden Potential of Cooling Towers]]></title>
            <link>https://medium.com/climate-conscious/noya-unlocking-the-hidden-potential-of-cooling-towers-5c165d9bb491?source=rss-6fed2e13225a------2</link>
            <guid isPermaLink="false">https://medium.com/p/5c165d9bb491</guid>
            <category><![CDATA[cooling-tower]]></category>
            <category><![CDATA[climate-solutions]]></category>
            <category><![CDATA[climate-change]]></category>
            <category><![CDATA[carbon-capture]]></category>
            <category><![CDATA[carbon-emissions]]></category>
            <dc:creator><![CDATA[Loni Olowookere]]></dc:creator>
            <pubDate>Wed, 02 Feb 2022 17:32:27 GMT</pubDate>
            <atom:updated>2022-02-02T17:32:27.907Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Noya carbon capture equipment installed on the roof" src="https://cdn-images-1.medium.com/max/1024/1*lqQoOxNWSg2NkbuVoDEOlA.jpeg" /><figcaption>Source: Noya Team</figcaption></figure><p><strong>Direct Air Capture </strong>(DAC) has rightfully earned a place as one of the foremost carbon removal techniques on the planet and this has given rise to the speculation that DAC technologies can be deployed and commercialized to curb global warming. Pioneer DAC corporations like <a href="https://climeworks.com/">Climeworks</a> and <a href="https://carbonengineering.com/">Carbon Engineering</a> have successfully launched full-scale operations capturing a few thousand tons of CO2 yearly. While this is a positive development, it remains a rather minimal effort, considering the more than <a href="https://www.globalwarmingindex.org/">2 trillion of CO2 already in the atmosphere</a> and the question on many minds — Is it pragmatic to invest hundreds of millions of dollars in DAC to capture only a few million tons of CO2 annually?</p><p>In an effective counter to these lofty considerations, <a href="https://www.noya.co/">Noya</a>, a San Francisco-based start-up has come up with an unconventional approach to reduce the upfront capital costs and time required to perform direct air capture. In a brief chat with <a href="https://www.linkedin.com/in/joshisantos/">Josh Santos</a>, co-founder, and CEO, we discussed the cost implications of Noya’s technology, market reception of the idea, utilization possibilities for sequestered carbon etc.</p><h3>Two Million Cooling Towers</h3><p>Cooling towers are more common than you think and they are used whenever an elaborate cooling system is required, either for residential or industrial purposes. There are more than <a href="https://www.nsf.org/news/road-map-improve-response-legionnaires-disease-outbreaks-conserve-water">2 million cooling towers</a> are operating in the United States today and they span across multiple sectors from dairy agriculture, manufacturing to commercial real estate, mining, etc. Though they come in different shapes and sizes, cooling towers typically circulate big streams of water and air to remove heat from a building. However, Noya’s improvisation to this process involves adding a chemical solution that binds water to CO2 as well as a plumbing tee that diverts the CO2-infused water at the bottom tub of the tower into an exterior processing unit that separates the carbon dioxide, storing it in pressurized cylinders.</p><figure><img alt="A visual image of Noya’s carbon capture process." src="https://cdn-images-1.medium.com/max/1024/1*28Q5xNcCO8OGFT9ZQQ89UA.png" /><figcaption>Source: Noya Team</figcaption></figure><p>If Noya’s retrofit system is implemented in the two million cooling towers currently operational in the United States, 10 billion tons of CO2 can be captured annually. For a point of comparison, the United States <a href="https://www.c2es.org/content/u-s-emissions/">emits 6–7 billion tons of all greenhouse gases</a> into the atmosphere every year. Using Noya’s retrofitting system for its two million cooling towers, the United States will be able to achieve carbon negativity sooner than imagined.</p><h3>Cost Implications &amp; Market Reception</h3><p>But who foots the bill?</p><p>Noya covers the total cost of retrofitting the cooling towers and also the cost of running the process, taking the CO2, and selling it for carbon credits. The proceeds from that sale are shared with host partners. It is indeed a symbiotic arrangement with Noya getting direct access to infrastructural facilities and host partners getting credits that they can utilize in their carbon accounting.</p><p>Due to the mutually beneficial nature of Noya’s approach, it is no surprise that market reception has been resoundingly positive so far. In Josh’s words, “Our approach has been well received so far and we have raised several rounds of funding, current investors include Fifty Years, LowerCarbon Capital and YCombinator. We are fortunate to have their backing and support. On the client-side, we have over a hundred locations that have signed up for our deployment waitlist and they are waiting for us to commence the technical implementation on top of their buildings”. Meanwhile, one deployment has been successfully executed and Noya’s scientists are working on a new version of the technology which they intend to scale up to the deployment phase within the next six months.</p><h3>Carbon Sequestration &amp; Utilization Pathways</h3><p>Currently, CO2 is used as an industrial condiment across traditional sectors like mining, food &amp; beverages, manufacturing etc and it is also becoming increasingly popular among startups who have the intention of utilizing it as an industrial feedstock to be repurposed into commodities like <a href="https://aetherdiamonds.com/">diamonds</a>, <a href="https://www.carboncraftdesign.com/carbontile">floor tiles</a>, <a href="https://www.pollima.co/">furniture</a> and even <a href="https://www.airprotein.com/">food</a>.</p><p>Noya is looking to partner with local sequestration partners for now while exploring the development of its sequestration technology to remove CO2 permanently in the future. The approaches being considered by Noya at the moment include the infusion of carbon into concrete or the use of captured carbon in depleted oil and gas mines. However, Josh admits that as they scale their cooling tower technology across the country, certain geographies will lend themselves better to other sequestration pathways different from underground storage in depleted mines because he is deliberately trying to steer the company away from collaborating with oil and gas companies.</p><h3>Carbon Capture Vs Emission Incentivization</h3><p>I asked Josh to weigh in on this ethical debate considering that artificial carbon capture approaches are arguably the <a href="https://www.axios.com/al-gore-technology-capturing-co2-emissions-nonsense-c1e5e230-0ffb-4b5c-ba24-8bcdfd78e7d3.html">most criticized</a> for several reasons including high deployment cost, possible greenwashing dalliance with fossil fuel companies and the incentivization of further emissions. In response, Josh made a distinction between post-combustion carbon capture and Noya’s cooling tower approach, stating, “I do not think our approach encourages further emissions as we are not doing post-combustion capture from smokestacks, the CO2 we capture is the cleanest source of CO2 available since it is gotten from the ambient atmosphere. Some of the companies that we are deploying our technology to do not directly emit anything outside of their scope asides from their connection to the local energy grid. So we are not furthering or encouraging the use of fossil fuels.”</p><p>He continues by emphasizing who the actual decision-makers in the energy sector are, “…I believe what is going to eventually drive us away from fossil fuels and make the biggest impact is the individual decisions that firms and consumers are making about where their energy comes from and what type of energy they are using. So, I think more than the carbon accounting of an organization that makes money from putting CO2 in the air, I think the way we get off of fossil fuels comes from the decisions we are making from the local and the state level.”</p><p>While the race to reduce global emission rates is important, t<a href="https://news.un.org/en/story/2020/12/1078612">he United Nations has also made it clear</a> that we need carbon removal technologies as well to pull CO2 out of the atmosphere to push global temperatures down to a level that is out of the danger zone before time runs out. So, with or without incentivization, capturing carbon is already an important aspect of keeping climate change reversible and if we need the technology to be effective at scale 20 years from now then we need to start working on it now.</p><h3>On Propaganda and the Climate Community at Large</h3><p>When asked if climate propaganda distracts from actual work in the community, Josh replied by saying, “I think it is important we all speak from the same set of understanding. The most important thing that we could do is talk about science, talk about the sources of the problem, sources of CO2 emissions, and how to reduce them. I think anything outside of those discussions could be a distraction even if it is well-intentioned. People need to understand why certain actions are required and the scientific basis for it. I think all kinds of climate communication should arise and stay grounded in scientific facts. It is quite simple, we have two levers to reverse climate change — reducing carbon emissions and removing the carbon already in the atmosphere. Anything outside of those two levers should be given a hard look at.”</p><h3>Future Outlook</h3><p>It is difficult to tell at this point the crucial role DAC technologies may eventually play in the next decade as every hand is being called on deck to reverse climate change. According to David Izikowitz, the founder of <a href="https://www.carboninfinity.com/">Carbon Infinity</a>, the successful adoption and deployment of DAC technologies to a scale that will significantly reduce the trillions of CO2 in the atmosphere will require “<a href="https://www.frontiersin.org/articles/10.3389/fclim.2021.636657/full">wartime-like mobilization of supply chains, manufacturing capacity, and innovation</a>”. However, companies like Noya that can take advantage of the existing infrastructural facilities to mitigate the installation costs of carbon capture technology will probably be safer bets in the long run.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5c165d9bb491" width="1" height="1" alt=""><hr><p><a href="https://medium.com/climate-conscious/noya-unlocking-the-hidden-potential-of-cooling-towers-5c165d9bb491">Noya: Unlocking the Hidden Potential of Cooling Towers</a> was originally published in <a href="https://medium.com/climate-conscious">Climate Conscious</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>