Graph ML for VC | Part 12 | Graph ML for Video Understanding | CVPR 2022 Tutorial

0:00 Introduction
4:01 Training Loss of GCNs with varying depth
5:09 Residual Graph Connections
5:59 Dilated Graph Convolutions
7:26 DeepGCNs for Node Prediction
8:44 Popular Open Source Codebase
9:17 Memory Complexity of Training Deep GNN
11:08 Results: Constant Memory with RevGNN
12:26 Video Understanding Tasks
13:20 Temporal Activity Localization T
13:59 Video is represented by clips
15:12 Video as a graph (for context)
15:30 Graph-Based TAL
17:20 G-TAD Pipeline
20:11 Context is adaptive with Dynamic GCNs
22:09 Graph Pyramid Network for TAL
22:45 Video-Language Grounding (VLG)
24:08 Video-Language Graph Matching Network VLG
26:02 MAD: Large-Scale VLG Dataset
26:26 Active Speaker Detection (ASD)
27:48 Multi-modal Assignation for ASD MAAS