The Large Scale Robotic Vision Perception Challenge

We propose to develop a large scale robotic vision benchmark to address the most critical challenges for robotic perception that are not covered by existing computer vision benchmarks: object detection in open-set conditions, incremental learning with low-shot techniques, active learning, and active vision. We will develop the performance metrics, benchmarking protocol, and evaluation server to measure the large-scale performance for these tasks.


Funding / Grants

  • Google Faculty Research Award $73k

Chief Investigators