Pulak Purkait
Papers
5
Total Citations
65
H-Index
3
About
Pulak Purkait’s research sits at the intersection of computer vision and robotics, with a focus on 3D localization, pose estimation, and semantic mapping. His most influential work, “SPP-Net: Deep Absolute Pose Regression with Synthetic Views” (38 citations), tackles the critical problem of image-based localization for autonomous systems and augmented reality by introducing a deep learning approach that regresses absolute camera pose from synthetic imagery. This contribution offers a scalable alternative to traditional geometric registration methods. Purkait has also made notable advances in planar marker pose estimation, developing robust rotation averaging techniques that resolve ambiguities in 6DOF pose estimation—a key challenge for mapping and localization in robotics. His work on dense RGB-D semantic mapping with a Pixel-Voxel neural network extends 3D mapping to include scene understanding, enabling robots to interpret both geometry and context. Across these contributions, Purkait demonstrates a consistent drive to enhance the reliability and intelligence of perception systems, bridging deep learning with geometric constraints for real-world deployment.
Research Focus
Key Achievements
Top Papers
- 1SPP-Net: Deep Absolute Pose Regression with Synthetic Views38 citations · 2017
- 2
- 3Dense RGB-D semantic mapping with Pixel-Voxel neural network4 citations · 2017
- 4
- 5NeuRoRA: Neural Robust Rotation Averaging2 citations · 2020