Project Description
Shadow detection and removal is one of the most basic and challenging problems in computer graphics and computer vision. Shadows not only affect the visual interpretation of the image, but also subsequent image processing steps. For instance, a darker area (caused by shadows) introduces incorrect segments in image segmentation, and radiation changes (caused by shadows) may reduce the performance of object recognition and tracking systems. Therefore, it is crucial to detect and remove shadows early in the processing pipeline.
The goal of this project is to develop and benchmark end-to-end shadow detection and removal systems. Students will work with their mentors to review and implement different systems for shadow removal and benchmark approaches based on both “classical” computer vision and deep learning techniques. Time permitting, students may explore practical applications of their computer vision system in autonomous driving, augmented reality, or visual accessibility.
Prerequisite Information
None required, but any of the following will be highly beneficial:
CS-165B (Machine Learning),
CS-180 (Computer Graphics),
CS-181 (Computer Vision - can also be taken concurrently with this project)
Knowledge/Skills to Acquire (with guidance from mentors)
Concepts: Image processing, computer vision, deep learning.
Programming: Keras, PyTorch
Team Members
- Alvin Wang
- Edward Ding
- Jennifer Zhu
- Kyle Zhao
Professor and Mentors
- Prof. Michael Beyeler
- Grad mentors: Apurv Varshney and Galen Pogoncheff
Meeting Times
- Mentor Meetings
- Thursdays, 3:15-3:45 p.m.
- ERSP Team Meetings
- Tuesdays, 9-11 a.m.