Shillin Hu , Ph.D. Research Proficiency Presentation: 'End-to-end orientation estimation from 2D cryo-EM images'

Tuesday, August 17, 2021 - 11:30am to 2:35pm
Zoom - contact for Zoom info.
Event Description: 
Title: Shadow detection for videos

Shadow understanding is one of the fundamental topics in computer vision tasks. While single image shadow detection has been improving rapidly in recent years, video shadow detection remains a challenging task due to data scarcity and the difficulty in modelling temporal consistency. The current video shadow detection method achieves this goal via co-attention, which mostly exploits information that is temporally coherent but is not robust in detecting fast appearance changing shadows and small shadow regions.

In this report, we will first review the existing shadow detection methods for single images and different ways to extend image-based methods to video-based methods on related topics. We then propose a simple but powerful method to better aggregate information temporally. An optical flow based warping module is used to align and then combine features between frames. We apply this warping module across multiple deep-network layers to retrieve information from neighboring frames including both local details and high-level semantic information. We train and test our framework on the ViSha dataset. Experimental results show that our model outperforms the state-of-the-art video shadow detection method by 28%, reducing BER from 16.7 to 12.0.

Hosted By: 
Dimitris Samaras
Computed Event Type: