The Computer Vision for Embedded Systems (CVES) team will investigate methods to improve efficiency (inference time, training time, storage space, energy consumption) of computer vision (both image and multimedia) so that computer vision can run on embedded systems. The team will evaluate how existing methods (such as quantization and pruning) can be applied to new neural architectures (such as transformers). The team will also investigate new architectures of neural networks and compare their efficiency with different levels of accuracy.

 

A student can select one of the following topics:

1. Design future Low-Power Computer Vision Challenges, collect data, establish rules, and create sample solutions; maintain and improve the submission website https://lpcv.ai/ and the evaluation systems
2. Analyze and improve past winning solutions of Low-Power Computer Vision Challenge.
3. Create education materials and programming assignments for future members to learn computer vision and embedded systems