<?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 Ram Ramrakhya on Medium]]></title>
        <description><![CDATA[Stories by Ram Ramrakhya on Medium]]></description>
        <link>https://medium.com/@ramramrakhya81?source=rss-816dd0a232c6------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*wSp_D9HJ0aYSB23djicfBA.jpeg</url>
            <title>Stories by Ram Ramrakhya on Medium</title>
            <link>https://medium.com/@ramramrakhya81?source=rss-816dd0a232c6------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 23:30:50 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ramramrakhya81/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[GSOC’18 with CloudCV : Wrap Up]]></title>
            <link>https://medium.com/@ramramrakhya81/gsoc18-with-cloudcv-3a238e1fc994?source=rss-816dd0a232c6------2</link>
            <guid isPermaLink="false">https://medium.com/p/3a238e1fc994</guid>
            <category><![CDATA[open-source]]></category>
            <dc:creator><![CDATA[Ram Ramrakhya]]></dc:creator>
            <pubDate>Fri, 10 Aug 2018 05:27:21 GMT</pubDate>
            <atom:updated>2018-08-13T05:21:49.367Z</atom:updated>
            <content:encoded><![CDATA[<p>For me the last 3 months were awesome, GSOC’18 has finally come to an end, I really enjoyed working on the project <a href="https://github.com/Cloud-CV/Fabrik">Fabrik</a>. The whole GSOC period has been a great experience starting from collaborating on project, submitting proposal till the completion of the project. It was a great learning experience which gave me an understanding of new perspective to view a problem for solving it. I would like to thank my mentors <a href="https://utsavgarg.github.io/">Utsav Garg</a>, <a href="https://github.com/kdexd">Karan Desai</a>, <a href="https://ayshrv.github.io/">Ayush Shrivastava</a>, <a href="http://deshraj.xyz/">Deshraj Yadav</a> at <a href="https://cloudcv.org/">CloudCV</a> for their support and guidance throughout the summer of code.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*g5RBYeGe0VLB6t_ZsvO_wQ.png" /></figure><h3>Project Goals</h3><p>The features that we decided to work on during the summers are given below:</p><ol><li>Extend support for tensorflow import and export</li><li>Finish setup for docker containers</li><li>Improve visualisation algorithm</li><li>Add support for real time model sharing to help users collaborate</li><li>Add support for open link model sharing</li><li>Add support for collaborators comments on individual layers of network</li><li>Add support for real time highlighting feature</li><li>Setup authentication system</li><li>Add support for celery to create asynchronous tasks</li><li>Create a homepage — work in progress</li></ol><h3>Contribution on Fabrik</h3><h4>Major Contributions</h4><p>Below is the list of contributions that I have made on EvalAI:</p><ol><li>PR <a href="https://github.com/Cloud-CV/Fabrik/pull/344">#344</a>: Complete docker container setup</li><li>PR <a href="https://github.com/Cloud-CV/Fabrik/pull/348">#348</a>, <a href="https://github.com/Cloud-CV/Fabrik/pull/349">#349</a>, <a href="https://github.com/Cloud-CV/Fabrik/pull/351">#351</a> : Improve visualisation algorithm.</li><li>PR <a href="https://github.com/Cloud-CV/Fabrik/pull/352">#352</a>, <a href="https://github.com/Cloud-CV/Fabrik/pull/353">#353</a>, <a href="https://github.com/Cloud-CV/Fabrik/pull/354">#354</a>, <a href="https://github.com/Cloud-CV/Fabrik/pull/350">#350</a>, <a href="https://github.com/Cloud-CV/Fabrik/pull/356">#356</a> : Extend tensorflow import and export support.</li><li>PR <a href="https://github.com/Cloud-CV/Fabrik/pull/357">#357</a> : Add setup for django channels.</li><li>PR <a href="https://github.com/Cloud-CV/Fabrik/pull/361">#361</a> : Add setup for celery.</li><li>PR <a href="https://github.com/Cloud-CV/Fabrik/pull/363">#363</a> : Add authentication support using github and google accounts.</li><li>PR <a href="https://github.com/Cloud-CV/Fabrik/pull/364">#364</a>: Real time collaboration feature for open link models.</li><li>PR <a href="https://github.com/Cloud-CV/Fabrik/pull/369">#369</a>: Real time commenting and model history feature.</li></ol><h4>Minor Contributions</h4><p>There are many minor contributions which includes bug fixes, minor UI changes etc.</p><p>Below is a list of hyperlinked Pull Request numbers:</p><p><a href="https://github.com/Cloud-CV/Fabrik/pull/345">#345</a>, <a href="https://github.com/Cloud-CV/Fabrik/pull/346">#346</a>, <a href="https://github.com/Cloud-CV/Fabrik/pull/355">#355</a>, <a href="https://github.com/Cloud-CV/Fabrik/pull/360">#360</a></p><h4>Still in review</h4><p>PR <a href="https://github.com/Cloud-CV/Fabrik/pull/370">#370</a>, <a href="https://github.com/Cloud-CV/Fabrik/pull/371">#371</a>, <a href="https://github.com/Cloud-CV/Fabrik/pull/373">#373</a></p><p><a href="https://docs.google.com/presentation/d/1HmOlxlHX1mduEgr_G-hFW5QT9TxDrBSvYel0ohCTStw/edit?usp=sharing"><strong>Detailed Overview</strong></a></p><h3>Conclusion</h3><p>During my GSoC project, I faced some challenges to find a balance between my job and GSoC tasks, but constant support and guidance from mentors helped me to balance both. Working with mentors has been a lot of fun, and also working with various people in the community was a great experience.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3a238e1fc994" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[GSOC’18 with CloudCV]]></title>
            <link>https://medium.com/@ramramrakhya81/gsoc18-with-cloudcv-a517d343e01f?source=rss-816dd0a232c6------2</link>
            <guid isPermaLink="false">https://medium.com/p/a517d343e01f</guid>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Ram Ramrakhya]]></dc:creator>
            <pubDate>Mon, 28 May 2018 12:51:23 GMT</pubDate>
            <atom:updated>2018-05-28T12:51:23.313Z</atom:updated>
            <content:encoded><![CDATA[<p>Getting shortlisted for GSOC’18 was the goal I was aiming since a year and on 23rd April 2018 results were out and guess what, I got shorlisted. These series of blogs will be based on the work I will be doing in my project during this summer of code.</p><h3>What is GSOC ?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Vx8L4x_-rZQ6zNEVjn_GcQ.png" /></figure><blockquote>To summarize Google Summer of Code is a global program focused on bringing more student developers into open source software development. Students work with an open source organization on a 3 month programming project during their break from school. You can read more about it <a href="https://summerofcode.withgoogle.com/">here</a>.</blockquote><h3>GSOC’18 Project — Fabrik</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/720/1*-RnwZcPP7gFeg0rbYbProA.png" /><figcaption>Collaboratively Build, Visualize, and Design Neural Nets.</figcaption></figure><p>For GSOC’18 I am working with CloudCV on the project Fabrik. Fabrik is an online collaborative platform to build, visualize and train deep learning models via a simple drag-and-drop interface. It allows researchers to collaboratively develop and debug models using a web GUI that supports importing, editing and exporting networks written in widely popular frameworks like Caffe, Keras, and TensorFlow.</p><p>My work during the period of summer of code will be focused on multiple features like:</p><ul><li>Extend support for Tensorflow Framework.</li><li>Adding Real Time Collaboration support for collaboratively building deep learning models.</li><li>Adding docker container support for setting up development environment.</li><li>Optimization of Visualization Algorithm.</li><li>Setting up Authentication mechanism.</li><li>Addition of chat interface for users to collaborate.</li></ul><h3>How does it work ?</h3><p>The work during the whole summer of code is divided into four phases and each phase has its set of tasks decided by mentors. There’s a timeline assigned to every student based on which the work during a particular phase is needed to be done. The four phases are as follows</p><ul><li>Phase 0 — Community Bonding Period</li><li>Phase 1</li><li>Phase 2</li><li>Phase 3</li></ul><h3>Phase 0 — Community Bonding Period</h3><p>The first three weeks from 23rd April, 2018 are part of <strong>Community Bonding Period. </strong>This is the time given to students to get familiar with Organization and mentors. In this phase of my project there were multiple tasks assigned which helped me get started with my timeline. The tasks completed during Phase 0 are as follows</p><ul><li><strong>Document the approach for Tensorflow Import</strong></li></ul><p>The first task was to document the approach for extending Tensorflow import support. The task focused on investigating the details of layers available in Tensorflow in order to understand the difficulties that could be faced while adding support for the layer, how would we handle the abstraction of layers so that layer definition is generalized for exporting into other frameworks, how would we handle the fine granularity in Tensorflow graph_def. The deliverable of the task was a document with possible layers for which support will be added with the difficulties (if any) and a rough estimate of time required for completing the feature.</p><ul><li><strong>Document the approach for Tensorflow Export</strong></li></ul><p>The second task was to document the approach for extending Tensorflow export support. The task focused on investigating the robustness of existing approach for Tensorflow export and the custom approach for building export functionality. The aim of this task was to compare pros and cons of both the approach and decide on the more robust solution for Tensorflow export functionality. The deliverable of the task was a document with both the approaches with their pros and cons.</p><ul><li><strong>Approach discussion for Real Time Collaboration</strong></li></ul><p>The third task was to decide the approach for Real Time Collaboration feature. The task focused on deciding on how to model states for implementing to apply updates, which type of architecture will be used, decision of technologies to work on, discussion on Operational Transformation to handle real time conflict resolution. The deliverable of task was a document with the approach for feature implementation.</p><ul><li><strong>Addition of docker container support for setting up development environment</strong></li></ul><p>The fourth task was to fix the issues in existing docker container support for setting up the development environment. The task focused on investigating the issues with the existing dockerized setup and fixing the issues to complete the docker container support. The deliverable of the task was a docker container support for setting up Fabrik development environment.</p><ul><li><strong>Optimization of Visualization Algorithm</strong></li></ul><p>The fifth task was to improve &amp; optimize existing Visualization Algorithm. The task focused on optimizing the existing logic of visualizing a Neural Network on the platform by reducing the redundant operations and optimization of existing Topological Sorting algorithm using <a href="https://en.wikipedia.org/wiki/Topological_sorting">Kahn’s Algorithm</a>. The other part of the task was to assign position to layers of a Neural Network in a such a way that model is cleaner to visualize for end user. The second part of task was based on devising a hashing based approach to find appropriate positions to layers while plotting on canvas. The deliverable of the task was a improvised visualization algorithm.</p><p>This was all for the <strong>Community Bonding Period. </strong>During the whole Phase - 0 of summer of code mentors guided me through doubts and issues related to tasks, they also helped improvising approaches and were open to suggestions.</p><p>There are some awesome features lined up for next phases, <strong>Stay Tuned!</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a517d343e01f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Google Summer of Code]]></title>
            <link>https://medium.com/@ramramrakhya81/google-summer-of-code-bc825439cb06?source=rss-816dd0a232c6------2</link>
            <guid isPermaLink="false">https://medium.com/p/bc825439cb06</guid>
            <category><![CDATA[programming]]></category>
            <dc:creator><![CDATA[Ram Ramrakhya]]></dc:creator>
            <pubDate>Wed, 07 Feb 2018 09:51:56 GMT</pubDate>
            <atom:updated>2018-05-07T06:32:07.693Z</atom:updated>
            <content:encoded><![CDATA[<p>In December 2016 I decided to participate in Google Summer of Code , first step in order to participate was I needed an organisation in which I would like to contribute, I was looking up in the list of past organisations so that I can decide where I would like to apply. While I was randomly scrolling down the page of past organisations I saw one organisation named “CloudCV”, when I saw the name of the organisation I thought that it would be some organisation which works on computer vision (I was biased for thinking like that just because I knew about OpenCV 😝), I saw their projects where the backend was in Django and the frontend was in AngularJS and ReactJS. After seeing those projects I was impressed with the kind of work they were doing and I decided to apply for CloudCV. Though I had no experience in any of those technologies they were using. 😛</p><p>I started with a project named EvalAI in December where I was working on some small issues and on January 1st I submitted my first Pull Request officially. In the month of December I learned a lot of stuff like git, AngularJS basics and Django basics. The reviewers of that organisation then started giving me suggestions to make the fix as per their guidelines. My first Pull Request being accepted was a different kind of experience, from the day I submitted the PR till the day it was accepted was fun in its own way. From that day onwards I started understanding the codebase of project and I was even bunking lectures to work on that project. It was a one of a kind experience writing solutions to the issues and submitting the pull requests for fixing the issue and waiting for it to get accepted by mentors. There were two more candidates who were working as hard as I was for the same project “EvalAI” for GSOC’ 2017 and soon I was one of top 2 contributors on the project among all the applicants for GSOC’ 2017.</p><p>Then came the time for submitting the proposal for the project you wanted to apply for the period of GSOC. I started with my proposal and when it was almost complete one of the mentor suggested me to apply for their other project rather than applying for “EvalAI”, CloudCV had a total 4 projects in GSOC’ 17. He asked me to apply to some other project recognising my effort towards working for the organisation. Then I switched to writing a proposal for “Fabrik” another more interesting project of CloudCV, but I didn’t have much experience with deep learning frameworks at that time, so I started understanding it and wrote my proposal and submitted it.</p><p>Then came the day of results for selection of candidates for GSOC’ 17 and I wasn’t one of the selected ones, another guy got selected for the project I applied for, because of my lack of experience in the domain of deep learning. But I didn’t give up and started learning, tried understanding various frameworks and working on my technical skills. While other candidates were completing their timeline for GSOC’ 17, I had already started preparing for GSOC’ 18.</p><h3>GSOC’ 18</h3><p>I started my work for GSOC’ 18 from September 2017 with Fabrik, I knew that if I want to do GSOC this is the project I want to work on. As a newbie in deep learning people face many issues while building a model of their own as they have to spend a long time going through the documentation of frameworks and languages (though that’s the most enjoying part of learning new stuff personally speaking :P). I started with my PR to the project by adding an example to Fabrik/Zoo (that’s what we call our model example store, similar to Caffe/Zoo). My first few PRs were focused on most basic stuff of the project, then as I was progressing slowly towards understanding codebase I decided to start with some moderate difficulty issue. The issue was focused on the main functionality of the project (which is obviously the visualisation of deep learning model graph on the canvas which looks similar to the flowchart), the task was to improvise the existing algorithm to avoid cluttered visualisation of graphs. I found an interesting solution to the problem by using a concept called “Hashing”, it felt really awesome when I was able to apply my knowledge of competitive coding to solve issues in the project.</p><p>After completing the task I submitted the PR which got accepted felt really good. As I said earlier I really enjoyed when my PR got accepted, which gave me confidence to start up with more complicated tasks. In this way I was able to solve few moderate difficulty issues and few hard issues which also gave me much more understanding of the codebase of project.</p><h3>The Proposal</h3><p>Then came the time for proposal submission, before I started with my proposal I had a rough picture in my mind about what all features should be added to the project, what all improvements need to be made to existing system, etc. I chalked out all the details, mentioned what all features I am planning to incorporate, how will I approach them, what all things need to be considered for implementing those features, a scheduled timeline that I’ve planned out for the work that needs to be done, how much time I’ll be dedicating to the project, what all new technologies I would require to learn. After completing my initial draft I shared it with the mentors and based on their feedback made the changes to the proposal.</p><h3>The Interview</h3><p>After submitting the proposal I was waiting for the day of results, when I was randomly checking my mailbox I saw that I received an interview invite from CloudCV for discussing my proposal further. I was really excited for the interview, the mentors began with discussion about the proposal they said that there are few changes in the deliverables required for the project and gave me a task to devise an approach to implement the feature of Real Time Collaboration (yup! similar to what we do in Google Docs just for a flowchart or you can say a deep learning model, sounds interesting right?). I thought on it for few minutes and gave my answer to them, they were satisfied with the approach but they asked me to research a bit more on the same and come up with a better approach. I took few days and gave them an improvised solution for the same.</p><h3>The Result</h3><p>Then came the day of results, I was eagerly waiting for the clock to turn time to 9:30 PM quickly with the page of organization open on summer of code portal, as soon as the clock hit 9:30 I refreshed the page and what do I see ? Yeah you guessed it right, I got selected for GSOC’ 18. It was one of the happiest moment, I was really happy that I’ll be spending my summer doing something I love (i.e. code :D).</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bc825439cb06" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>