Srijan Das

About

Srijan Das is a rising researcher at the intersection of computer vision and reinforcement learning, with a focus on making AI systems more perceptive and efficient. His work centers on two key areas: enhancing reinforcement learning from visual inputs and improving how vision transformers understand human poses. In his highly cited 2022 paper, *"Does Self-supervised Learning Really Improve Reinforcement Learning from Pixels?"* (9 citations), Das critically examines the widely adopted contrastive reinforcement learning framework (e.g., CURL), conducting extensive experiments to determine whether self-supervised learning truly boosts online RL from pixel data—a foundational contribution to the field. Building on this, his 2023 work, *"Seeing the Pose in the Pixels: Learning Pose-Aware Representations in Vision Transformers"* (4 citations), tackles a critical gap in computer vision. Das proposes novel methods to embed pose awareness directly into Vision Transformers, enabling them to better capture human skeletons and robotic arm configurations. This breakthrough has direct implications for human action recognition and robot imitation learning. By bridging self-supervised learning, pose estimation, and reinforcement learning, Das is shaping how machines perceive and interact with dynamic environments.

Research Focus

Key Achievements

Top Papers

1. 1
   Does Self-supervised Learning Really Improve Reinforcement Learning from Pixels?

   9 citations · 2022
2. 2
   Seeing the Pose in the Pixels: Learning Pose-Aware Representations in Vision Transformers

   4 citations · 2023

Key Collaborators

Contact & Links