Jishnu Jaykumar P
Papers
6
Total Citations
32
H-Index
4
About
Jishnu Jaykumar P is a rising researcher at the intersection of computer vision, robotics, and few-shot learning. His work centers on enabling machines to recognize and manipulate objects with minimal training data—a critical capability for real-world robotics. His most impactful contribution is **Proto-CLIP**, a novel framework that fuses vision-language models like CLIP with prototypical networks, achieving strong few-shot learning performance by leveraging both image and text prototypes (9 citations). To address the scarcity of suitable benchmarks, he created the **FewSOL dataset**, a rich collection of 336 real-world objects with RGB-D images, segmentation masks, poses, and attributes, specifically designed for few-shot object learning in robotic environments (5 citations). He also introduced **SceneReplica**, a reproducible real-world benchmark for evaluating robot pick-and-place tasks using the standard YCB object set (4 citations). Most recently, his work on **NIDS-Net** tackles the challenging problem of novel instance detection and segmentation, offering a unified framework for identifying unseen objects from just a few examples. Through these contributions, Jishnu is building the foundational tools and datasets that will enable more adaptive, data-efficient robotic systems.
Research Focus
Key Achievements
Top Papers
- 1Proto-CLIP: Vision-Language Prototypical Network for Few-Shot Learning9 citations · 2024
- 2Proto-CLIP: Vision-Language Prototypical Network for Few-Shot Learning9 citations · 2023
- 3FewSOL: A Dataset for Few-Shot Object Learning in Robotic Environments5 citations · 2023
- 4
- 5FewSOL: A Dataset for Few-Shot Object Learning in Robotic Environments3 citations · 2022
- 6