About Me
2nd Year UG student with unsatiated hunger to advance in the field of deep learning intersecting with computer vision and computer graphics, research and specialize in 3D reconstuction and view synthesis. Finds zen staring at decrementing loss counters while listening to korn.
Oct 2023 - Present
Chennai, IN
- Contributed to a portfolio of multiple Deep Learning projects, gaining proficiency in Computer Vision and Computer Graphics disciplines.
- Trained many models utilizing frameworks such as PyTorch Lightning and Weights & Biases for experiment tracking, enhancing project efficiency and reproducibility.
Oct 2023 - Present
Chennai, IN
Curated an article on DataX Journal exploring the revolution of the Transformer Architecture in the 2020s.
Finetuned a DistilBERT model to detect toxic behaviour online, enhancing child safety in digital environments, achieved an F1-score of 0.97 on testing.
Mar 2024 - Apr 2024
Implemented the research paper ”NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis” and trained the NeRF on a synthetic NeRF dataset with 400px resolution.
Leveraged Lightning for a flexible and scalable pipeline and W&B for tracking over 10+ experiments.
Achieved an average Peak Signal-to-Noise Ratio score of 29.20 on the test set for View Synthesis.
Feb 2024 – Mar 2024
Developed a recursive ray tracing algorithm to render scenes with photo-realistic lighting and shading effects.
Leveraged the GLM library for streamlined vector and matrix computations, optimizing computational performance and enhancing code readability.
Rendered 7 scenes using the ray tracer, showcasing a diverse range of visual compositions.
Jan 2024
Utilized SIFT algorithm to extract robust keypoints and descriptors from overlapping regions of each image pair.
Employed RANSAC to estimate the homography matrix, defining geometric transformations between image pairs.
Introduced weighted blending techniques to ensure smooth transitions, reducing artifacts by 50%.
Orchestrated recursive stitching of intermediate panoramas to optimize render efficiency.
Aug 2023 – Nov 2023
Implemented the ViT-B/16 model from the research paper ”An image is worth 16x16 words” in PyTorch.
Analyzed the motivations behind the architecture and conducted image classification experiments.
Authored a detailed blog post on Medium exploring Transformers and their social impacts. Garnered over 50 likes within the 1st week of publication.