HyperAI超神経

プラットフォーム

Action Detection On Charades

評価指標

mAP

評価結果

このベンチマークにおける各モデルのパフォーマンス結果

		Paper Title
TTM	28.79	Token Turing Machines
CTRN	27.8	CTRN: Class-Temporal Relational Network for Action Detection
Coarse-Fine Networks (w/ self-supervised detection pretraining)	26.95	Weakly-guided Self-supervised Pretraining for Temporal Activity Detection
UniMD+Sync. (RGB+Flow)	26.53	UniMD: Towards Unifying Moment Retrieval and Temporal Action Detection
PDAN (RGB+Flow)	26.5	PDAN: Pyramid Dilated Attention Network for Action Detection
PAT	26.5	PAT: Position-Aware Transformer for Dense Multi-Label Action Detection
MS-TCT (RGB only)	25.4	MS-TCT: Multi-Scale Temporal ConvTransformer for Action Detection
3D ResNet-50 + super-events pretrained on AViD	25.2	AViD Dataset: Anonymized Videos from Diverse Countries
Coarse-Fine Networks	25.1	Coarse-Fine Networks for Temporal Activity Detection in Videos
MLAD (RGB + Flow)	23.7	Modeling Multi-Label Action Dependencies for Temporal Action Localization
I3D + biGRU + VS-ST-MPNN	23.7	Representation Learning on Visual-Symbolic Graphs for Video Understanding
3D ResNet-50 pretrained on AViD	23.2	AViD Dataset: Anonymized Videos from Diverse Countries
TGM (RGB+Flow)	22.3	Temporal Gaussian Mixture Layer for Videos
Super-events (RGB+Flow)	19.41	Learning Latent Super-Events to Detect Multiple Activities in Videos
R-C3D	12.4	R-C3D: Region Convolutional 3D Network for Temporal Activity Detection
Sigurdsson et al.	9.6	Asynchronous Temporal Fields for Action Recognition

0 of 16 row(s) selected.