BSN: Boundary Sensitive Network for Temporal Action Proposal Generation

ECCV 2018 • Tianwei Lin • Xu Zhao • Haisheng Su • Chongjing Wang • Ming Yang

Temporal action proposal generation is an important yet challenging problem, since temporal proposals with rich action content are indispensable for analysing real-world videos with long duration and high proportion irrelevant content. This problem requires methods not only generating proposals with precise temporal boundaries, but also retrieving proposals to cover truth action instances with high recall and high overlap using relatively fewer proposals... (read more)

PDF Abstract ECCV 2018 PDF ECCV 2018 Abstract
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT BENCHMARK
Temporal Action Localization ActivityNet-1.3 BSN mAP IOU@0.5 46.45 # 8
mAP 30.03 # 9
mAP IOU@0.75 29.96 # 8
mAP IOU@0.95 8.02 # 6
Temporal Action Proposal Generation ActivityNet-1.3 BSN AUC (val) 66.17 # 5
AR@100 74.16 # 5
AUC (test) 66.26 # 1
Temporal Action Proposal Generation THUMOS' 14 BSN + Soft-NMS AR@100 46.06 # 1
AR@1000 64.52 # 1
AR@200 53.21 # 1
AR@50 37.46 # 1
AR@500 60.64 # 1
Action Recognition THUMOS’14 BSN mAP@0.3 53.5 # 3
mAP@0.4 45.0 # 3
mAP@0.5 36.9 # 3
Temporal Action Localization THUMOS’14 BSN UNet mAP IOU@0.5 36.9 # 7
mAP IOU@0.3 53.5 # 5
mAP IOU@0.4 45 # 6
mAP IOU@0.6 28.4 # 5
mAP IOU@0.7 20 # 4

Methods used in the Paper


METHOD TYPE
🤖 No Methods Found Help the community by adding them if they're not listed; e.g. Deep Residual Learning for Image Recognition uses ResNet