Boundary Content Graph Neural Network (BC‑GNN) for Temporal Action Proposal Generation

The Boundary Content Graph Neural Network (BC‑GNN) introduces a bipartite‑graph framework that jointly refines start/end boundary probabilities and segment‑content confidence, enabling more precise temporal action proposals and achieving state‑of‑the‑art results on ActivityNet‑1.3 and THUMOS14.

BC-GNNComputer VisionDeep Learning

0 likes · 10 min read

Boundary Content Graph Neural Network (BC‑GNN) for Temporal Action Proposal Generation

ITPUB

Aug 7, 2020 · Artificial Intelligence

How BC‑GNN Improves Temporal Action Proposals with Boundary‑Content Graph Modeling

The paper introduces Boundary Content Graph Neural Network (BC‑GNN), a graph‑based approach that jointly models boundary and content predictions to generate more accurate temporal action proposals and reliable confidence scores, achieving state‑of‑the‑art results on ActivityNet‑1.3 and THUMOS‑14.

BC-GNNECCV2020Graph Neural Network

0 likes · 12 min read

How BC‑GNN Improves Temporal Action Proposals with Boundary‑Content Graph Modeling