This repo contains the Scene Separation and Data Selection Algorithm used for real-time video stream analysis.
This algorithm is a work featured in Spatio-Temporal Reasoning and Learning 2022.
Go to /slides and /workshop-paper to see the workshop presentation slides and paper.
- Refine 2SDS architecture
- consider a graph neural network approach - See the details in the slides
- Looking ahead - consider the impotency when GPT4 tries to process a video or image sequence? Can we integrate or enhance our methods to change the situation?
