Conducting in-depth research on image segmentation models for a paper to be submitted to CVPR 2024.Analyzed and Implemented CLIP, Dense-CLIP, CAT-Seg, and SAM: Thoroughly studied the original papers to understand their architectures and debugged the code by identifying and resolving issues such as incorrect model parameters, runtime exceptions and data preprocessing error through systematic testing and verificationIntegration of SAM Image Encoder into CAT-Seg: Investigated the compatibility of SAM’s image encoder with the CAT-Seg architecture and fine-tuned the integrated model on a relevant dataset to optimize the performance.Experimental Analysis and Debugging: Performed experiments to analyze the performance on various datasets for testing and debugged issues related to model convergence, overfitting and computational inefficiencies.

Implemented and debugged codes for various domain adaptation methods, including original papers like CoOp, CoCoOp, PointCLIPv2, DAPL, DANN and fixed issues like runtime error and data preprocessing errorConducting extensive research on domain adaptation techniques and their applications in computer vision.Preparing a paper on these implementations and findings for better results for submission to CVPR 2024.

• Successfully managing the academic course assisting the professors course having 750+ enrolled students• Conducting tutorial classes, doubt sessions, exam sessions and organising the whole TA team as a Head TA

Successfully coordinated and managed the event having 500+ athletes for the execution of Half-Marathon

Successfully mentored two groups of students participating in the hackathon in a collaborative platform