DSpace Collection:https://dspace.iiti.ac.in:8080/jspui/handle/123456789/96582026-05-12T09:27:44Z2026-05-12T09:27:44ZDesign of CNN for multi-class classification of videos and imagesKatewa, VineshRoopraj, B SJain, NamanTiwari, Aruna [Guide]https://dspace.iiti.ac.in:8080/jspui/handle/123456789/104112025-05-30T06:38:08Z2022-05-26T00:00:00ZTitle: Design of CNN for multi-class classification of videos and images
Authors: Katewa, Vinesh; Roopraj, B S; Jain, Naman; Tiwari, Aruna [Guide]
Abstract: Classification of images and action in a video are a very challenging task requiring high
computation resources. Image classification in itself is a complex task where multiple
convolution and pooling layers extract features from the images and these features go
through a set of fully connected layers that classify the features extracted into different classes, this task becomes even more complex and resource heavy with the addition of
another dimension i.e. time for video classification problems. Models like VGG, ResNet
and NASNetLarge are one of the go-to models for image classification as they provide
great results with the complexity of the model that they bring. Using the ResNet and
NASNetLarge we proposed a new model for image classification that takes the best from
ResNet and NASNetLarge structure, the proposed model is essentially a ResNet50 model
with better feature extraction capability from NASNetLarge.
For the task of Video classification we went through a number of models and their
combination with each other, SlowFast network is a model that performs 3D convolution
on the video frames with different frame rates and combines them to get the class
of the input. SlowFast uses ResNet50 model as it’s base so we simply used the
Modified architecture from image classification here. Another model that we reviewed
was VideoCapsuleNet and U-Net, combined these models as well to create a better
video classification model. Finally, a SlowFast model implemented in PyTorch with
modification from U-Net.2022-05-26T00:00:00ZBoolean satisfiabilityBellamkonda, RohithChaudhari, Narendra S [Guide]https://dspace.iiti.ac.in:8080/jspui/handle/123456789/104102025-05-30T06:38:08Z2022-05-26T00:00:00ZTitle: Boolean satisfiability
Authors: Bellamkonda, Rohith; Chaudhari, Narendra S [Guide]
Abstract: I have tried to implement an SAT solver using the Davis–Putnam–Logemann– Loveland (DPLL) algorithm. The DPLL algorithm consists of three stages, Unit Propagation, Pure literal Elimination and backtracking by choosing a literal which occurs most frequently in all the clauses combined. I used ReactJS to make the user Interface for inputting the clauses and the internal code takes the input and performs the unit propagation, pure literal Elimination and backtracking and provides the output stating whether the given Boolean formula in conjunctive normal form is satisfiable (there exists a result of true for various combination of Boolean values) or unsatisfiable (when various combination of Boolean values of the given literal fails to provide a true value as a result)2022-05-26T00:00:00ZOptimizing operations involved in cell architectureNikam, RohitAhuja, Kapil [Guide]Sharma, Rohit [Guide]https://dspace.iiti.ac.in:8080/jspui/handle/123456789/104092025-05-30T06:38:08Z2022-05-25T00:00:00ZTitle: Optimizing operations involved in cell architecture
Authors: Nikam, Rohit; Ahuja, Kapil [Guide]; Sharma, Rohit [Guide]
Abstract: Teams use inbuilt-monitors to cut tickets whenever their service metrics are breaching thresholds. There
are separate alarms for each service metric for each cell. Whenever there is a regional issue (like an AWS
LSE (Large Scale Event)), teams get a lot of tickets for their cell services having the same root-cause.
Whenever there is a regional issue (like an AWS LSE (Large Scale Events)), teams get a multiple of tickets
for their services with the same root-cause. This causes two problems:
1. During LSEs, it can lead to a lot of noise in the ticket queue and make focusing difficult for the on call.
2. All these tickets have to be closed manually. It is also possible to miss some tickets.
We need to reduce the number of tickets corresponding to the same issue. As the number of silos per region
increases, this will become more important.
De-duping solutions should ideally have all these properties
1. We should reduce ticket count as much as possible
2. We should have one root cause per ticket
3. Implementation should be purely in terms of Carnaval (we’ll avoid introducing DJS jobs/services/etc. to
avoid new points of failure)2022-05-25T00:00:00ZAnalysing emerging algorithms in multi-agent reinforcement learningGutgutia, Yash VardhanAhuja, Kapil [Guide]https://dspace.iiti.ac.in:8080/jspui/handle/123456789/104082025-05-30T06:38:08Z2022-05-25T00:00:00ZTitle: Analysing emerging algorithms in multi-agent reinforcement learning
Authors: Gutgutia, Yash Vardhan; Ahuja, Kapil [Guide]
Abstract: Reinforcement Learning has witnessed significant advancement in solving various decision-making problems in Machine Learning (ML), most of which involve more than one agent. We categorize such problems as multi-agent problems and utilize Multi-Agent Reinforcement Learn ing (MARL) to solve them. In this project, we shall have a look at a family of multi-agent environments (PettingZoo) and analyze two state-of-the-art multi-agent algorithms - Multi-Agent Deep Deterministic Policy Gradient (MADDPG) and Deep Deterministic Policy Gradient (DDPG).
We aim to train the agents in these newly developed multi-agent environments under both
algorithms. After that, we shall analyze their training curves using appropriate benchmarking techniques and re-establish how MADDPG’s centralized critic plays an essential role in communication/coordination-based agents.2022-05-25T00:00:00Z