https://lazyCodes7.github.io/ Recent content on Rishab Mudliar Hugo -- gohugo.io en-US Copyright © 2026, Rishab Mudliar. Sat, 17 Feb 2024 12:09:56 +0530 https://lazyCodes7.github.io/blogs/cs236_6/ Sat, 17 Feb 2024 12:09:56 +0530 https://lazyCodes7.github.io/blogs/cs236_6/ About These blogs are my notes that represent my interpretation of the CS236 course taught by Stefano. Continuing.. A variational approximation to the posterior For cases when p(z|x; θ) is intractable we come up with a tractable approximation q(z; ϕ) that is as close as possible to p(z|x; θ) In the case of an image having it’s top half unknown this is what we came up with the last time https://lazyCodes7.github.io/blogs/cs236_5/ Sun, 11 Feb 2024 12:48:56 +0530 https://lazyCodes7.github.io/blogs/cs236_5/ About These blogs are my notes that represent my interpretation of the CS236 course taught by Stefano. Recap Autoregressive models: Chain rule based factorization is fully general Compact representation via conditional independence and/or neural parameterizations Autoregressive models Pros: Easy to evaluate likelihoods Easy to train Cons: Requires an ordering Generation is sequential Cannot learn features in an unsupervised way Latent Variable Models: Introduction Lots of variability in images x due to gender, eye color, hair color, pose, etc. https://lazyCodes7.github.io/blogs/cs236_4/ Mon, 05 Feb 2024 11:27:56 +0530 https://lazyCodes7.github.io/blogs/cs236_4/ About These blogs are my notes that represent my interpretation of the CS236 course taught by Stefano. Recap Learning a generative model. We are given a training set of examples, e.g., images of dogs We want to learn a probability distribution p(x) over images x such that Generation: If we sample xnew ∼ p(x), xnew should look like a dog (sampling) Density estimation: p(x) should be high if x looks like a dog, and low otherwise (anomaly detection) Unsupervised representation learning: We should be able to learn what these images have in common, e. https://lazyCodes7.github.io/blogs/cs236_3/ Fri, 26 Jan 2024 15:27:56 +0530 https://lazyCodes7.github.io/blogs/cs236_3/ About These blogs are my notes that represent my interpretation of the CS236 course taught by Stefano. Recap: Learning a generative model Recall that we want to learn a probability distribution p(x) over images x such that sampling from this distribution gives us new images. In the last part we dived deeper into probability distributions. Finally we saw ways to learn a probability distrubution for a discriminative model using techniques like logistic regression or neural models. https://lazyCodes7.github.io/blogs/cs236_2/ Wed, 24 Jan 2024 15:27:56 +0530 https://lazyCodes7.github.io/blogs/cs236_2/ About These blogs are my notes that represent my interpretation of the CS236 course taught by Stefano. Part 2 Learning a generative model In the last part we defined a statistical generative model represented as p(x) that lets us sample from this distribution to generate new images. But p(x) is unknown. How do we learn it? Before I go into it, we would be going into a lot of probablity terminologies that we will see. https://lazyCodes7.github.io/blogs/cs236_1/ Fri, 19 Jan 2024 15:27:56 +0530 https://lazyCodes7.github.io/blogs/cs236_1/ About These blogs are my notes that represent my interpretation of the CS236 course taught by Stefano. Part 1 Generative Modelling Generative modelling is a way to model/sample data that looks like something we might know. I am going to put more weight on the last part as it is important to generate something that makes sense. For instance, a machine learning model that generates ‘qweefejfww’ (random jargon) will not be of any use to us. https://lazyCodes7.github.io/posts/experience/ Fri, 27 Jan 2023 14:56:36 +0530 https://lazyCodes7.github.io/posts/experience/ Why am I writing this blog? Well I did my Google Summer of Code in 2022 that was like two years back. But every year I get a lot of people on LinkedIn or Twitter DMing about guidance, tips or just wanting to know about my experience. So this is a blog for them and to answer their questions! So let’s get started.. Why RedHenLab? Here in my college people are passionate to do GSoC like I am not kidding. https://lazyCodes7.github.io/posts/week16/ Fri, 09 Sep 2022 23:56:36 +0530 https://lazyCodes7.github.io/posts/week16/ This week Deployed the pipeline to Case HPC Documented and commented the codebase Current progress of the pipeline Stage 1: Curation After working on collecting paintings related to Christian Iconography for 3-4 weeks. The final version of the dataset consists of approximately 9.6k data points. Museums Web Gallery of Art - 5k images Art Institute of Chicago - 500-700 Met Art - 500-700 Corpus Viteraeum - 2k National Gallery of Art, Washington DC - 1k-2k Artforms. https://lazyCodes7.github.io/posts/week15/ Fri, 02 Sep 2022 23:56:36 +0530 https://lazyCodes7.github.io/posts/week15/ This week Implemented the transformer model for captioning task Presented the demo of the pipeline to the mentors Changing the dataset(torch.Dataset instance i.e) Earlier the dataset instance of PyTorch that I was using was simple. Whenever I referred an index it would return the image and numericalized version of the caption. But I wanted to use GPT2 for what it is good for. Hint: Text generation. So I changed the captions by providing a prompt as a token. https://lazyCodes7.github.io/posts/week14/ Thu, 25 Aug 2022 23:56:36 +0530 https://lazyCodes7.github.io/posts/week14/ This week Used a transformer based model to work on the captioning task A Transformer that uses images + text During the last week’s meeting I had concluded that using models like Word2Vec might not always work as they have a fixed representation of each word but words can differ contextually so this week I decided to use some better language models like GPT2 and BERT. VisionEncoderDecoder For incorporating transformer I used huggingface’s VisionEncoderDecoder implementation which helps us in initializing any vision-transformer(ViT, BEiT etc) based model as the encoder and a language model(GPT2, BERT, RoBERTa etc) as the decoder. https://lazyCodes7.github.io/posts/week13/ Thu, 18 Aug 2022 23:56:36 +0530 https://lazyCodes7.github.io/posts/week13/ This week Worked on using the captions I had collected during the curation stage I also build up a CNN+LSTM model to see if we could achieve good captioning results without use of Transformers Approach towards using the text. A image captioning model makes use of both text and an image but interpreting text can be difficult. Normally to solve this we use word embeddings which give us an idea about the similarity between two or more words. https://lazyCodes7.github.io/posts/week12/ Tue, 09 Aug 2022 23:56:36 +0530 https://lazyCodes7.github.io/posts/week12/ This week Read a chapter exclusively on saints from The Gothic Image by Emile Male. Annotated 200 images + previous images with more attributes I learned from reading the book by Emile Male Paintings in context This whole time I had been curating images and trying to annotate them but after reading 70 pages of this book. It made things really clear for me so I think it will be great to explain how Christian Paintings really work. https://lazyCodes7.github.io/posts/week11/ Thu, 04 Aug 2022 23:56:36 +0530 https://lazyCodes7.github.io/posts/week11/ This week Worked on moving my module to CaseHPC Tried perfecting F-RCNN for detecting saints and their attributes in paintings. Regarding CaseHPC One of the things I couldn’t do in the first phase of this program was to use Case Western’s HPC to train my models in general. So this week I decided to get my hands dirty and made my module work on one of the gpu nodes of Case HPC https://lazyCodes7.github.io/posts/week10/ Tue, 26 Jul 2022 23:56:36 +0530 https://lazyCodes7.github.io/posts/week10/ This week This week I completed the module that extracts various attributes from the paintings. Apart from that I will be showcasing in this blog whatever I have done in phase-1 The Modified Pipeline 1. Data Curation. 2. Image Captioning Model Changes I have made very minute changes to the pipeline. One of my goals was to create a module for adding more data points but I have kept that on hold. https://lazyCodes7.github.io/posts/week9/ Tue, 26 Jul 2022 14:56:36 +0530 https://lazyCodes7.github.io/posts/week9/ This week This week I spent my time working on the module that would extract features from paintings Meeting 1 Since I was done with my curation I wanted some idea on how to proceed forward with my project and so I had mailed my mentors regarding the next steps for this project. This meeting was mostly oriented with what I had done and what are the results I could produce for the first evaluation. https://lazyCodes7.github.io/posts/week8/ Mon, 18 Jul 2022 14:56:36 +0530 https://lazyCodes7.github.io/posts/week8/ This week This week I spent curating data from two sources. 1. Corpus Vitrearum. Corpus Vitrearum is a museum that has around 7000 stained glass artworks that are stored according to their iconclasses. By using various scraping techniques I scraped 2000 images that pertain to Christian iconographic content. During the curation process I am also logging some of the attributes from the metadata. The advantage of this is that it keeps the dataset open to more usecases. https://lazyCodes7.github.io/posts/week7/ Mon, 11 Jul 2022 14:56:36 +0530 https://lazyCodes7.github.io/posts/week7/ This week This week I worked on curating data points from two sources i.e the Art Institute of Chicago and the Metropolitan Museum of Art. 1. Art Institute of Chicago. The Art Institute of Chicago provides open access to around 50,000 artworks with rich metadata involved. By using the search tool provided by the Institute we have got around 1500 results that pertain to containing Christian content in them. Let’s look at one of the art pieces. https://lazyCodes7.github.io/posts/week5/ Sun, 03 Jul 2022 14:56:36 +0530 https://lazyCodes7.github.io/posts/week5/ This week I continued my search for licensed souces and also found a really good site that provides API access/open access to museum metadata After compiling enough sources for 2 weeks I finally started the process for curation. Source 1 - The Web Gallery of Art The Web Gallery of Art provides access to 60,000 artworks in the form of a dataset. This source has also been used in other art-related datasets like SemArt and ArtDL. https://lazyCodes7.github.io/posts/week4/ Sun, 19 Jun 2022 14:56:36 +0530 https://lazyCodes7.github.io/posts/week4/ This week The project kickoff Before the coding period began I had previously been working on classification of the Christian icons in the painting. But in order to caption art, we needed a good source. So I had been collecting sources here. In order to get some idea on how to proceed with the sources collected I contacted my mentor Fred. Minutes of the meeting. Attendees - Fred, Marcelo, Rishab, Tiago https://lazyCodes7.github.io/posts/week3/ Sun, 12 Jun 2022 14:56:36 +0530 https://lazyCodes7.github.io/posts/week3/ This week This week was mostly aimed towards meeting people from RedHenLab as well as our peers. Meeting #1 Professor Mark organized a meeting consisting of mentors and contributors. The contributors were limited to the people Mark is mentoring. It was a normal discussion where people talked about the things they are doing where they are from etc. Meeting #2 During the first meeting, Mark had informed me that Parth’s project and mine arise from a similar arena i. https://lazyCodes7.github.io/posts/week2/ Sat, 04 Jun 2022 14:56:36 +0530 https://lazyCodes7.github.io/posts/week2/ This week CWRU Accounts The first thing that happened this week was that we got our CWRU Gmail accounts. HPC Setup After this it was time to setup CWRU’s HPC(High Performing Cluster). The steps for that are following. First setup CWRU’s VPN by following the guide here. Make sure that Duo is setup as a 2FA option as each time a connection to the VPN occurs a subsequent push notification is sent to Duo Mobile. https://lazyCodes7.github.io/posts/week1/ Fri, 27 May 2022 14:56:36 +0530 https://lazyCodes7.github.io/posts/week1/ Introduction Hi new readers! I am Rishab Mudliar and I am currently in the final year of my Computer Science Major. This blog is part of a series dedicated to noting down my activities during the period of Google Summer of Code. GSoC Google Summer of Code or the acronym GSoC is a program aimed at people in tech looking to contribute to open-source software development. Selected contributors work for 12+ weeks on projects that solve interesting problems. https://lazyCodes7.github.io/archive/bertelsmann-scholarship/ Tue, 23 Mar 2021 15:27:56 +0530 https://lazyCodes7.github.io/archive/bertelsmann-scholarship/ Today I wanted to talk about my experience about the “Bertelsmann Scholarship” via this post. I hope after reading this, more people will be interested in what I believe is a truly amazing program for learning something that might interest you! Beginnings It all started when one of my friends recommending this to program back in November. The application had 3 choices “AI” or “Data” or “Cloud”. I decided to go with AI as I always wanted to try it out but didn’t have much knowledge about it. https://lazyCodes7.github.io/archive/networking-protocols/ Sat, 28 Mar 2020 14:56:36 +0530 https://lazyCodes7.github.io/archive/networking-protocols/ Protocols. So what exactly are protocols? Protocols according to me are certain set of rules that we use to communicate with each other Why do we use Protocols ? So in brief I explained about protocols but the next question is why do we need them.Let’s consider an example,say I am Thor and I wanna talk to some new avenger(Night Monkey) then first I need Night Monkey’s permission to talk(although I can just smash him with thunder and make him talk) and then Night Monkey will talk to Thor and so on a talk session starts. https://lazyCodes7.github.io/archive/first/ Wed, 01 Jan 2020 14:27:35 +0530 https://lazyCodes7.github.io/archive/first/ My First Blog! Hi there! I am Rishab Mudliar and coincidentally i am writing my first blog post on the first day of this year (I know this is a bad joke). I don’t know how bad this will be but i hope it’s not that bad that nobody understands.. Why Ubuntu? 1. Ubuntu is user-friendly Many computer users consider Linux-based systems hard to use and made for developers. It’s a huge misconception and Ubuntu Linux acts as a perfect myth-buster. https://lazyCodes7.github.io/projects/ Mon, 01 Jan 0001 00:00:00 +0000 https://lazyCodes7.github.io/projects/ Projects. CoVID Related. ShieldBot Shield Bot curates tweets about vital resources like ICU beds, plasma oxygen and more You can contact the supplier through the links that the bot sends you Twitter MaskAppeal A computer vision based script that detects specific facial features from a video feed. It determines the of the person/persons in the video frame is wearing a face mask or not within milliseconds for each frame input and requires minimal computational resources for each iteration.