Home Console Docs News Papers Tutorials Datasets Wiki SOTA LLM Models GPU Leaderboard Events

English

Action Detection On J Hmdb

Metrics

Frame-mAP 0.5

Video-mAP 0.2

Video-mAP 0.5

Results

Performance results of various models on this benchmark

				Paper Title	Repository
HIT	83.8	89.7	88.1	Holistic Interaction Transformer Network for Action Detection
HISAN (ResNet-101 + FPN)	-	87.59	86.49	Hierarchical Self-Attention Network for Action Localization in Videos	-
YOWO + LFB	75.7	88.3	85.9	You Only Watch Once: A Unified CNN Architecture for Real-Time Spatiotemporal Action Localization
YOWO	74.4	87.8	85.7	You Only Watch Once: A Unified CNN Architecture for Real-Time Spatiotemporal Action Localization
HISAN (VGG-16)	76.72	85.97	84.02	Hierarchical Self-Attention Network for Action Localization in Videos	-
MOC	74	80.7	80.5	Actions as Moving Points
Faster-RCNN + two-stream I3D conv	73.3	-	78.6	AVA: A Video Dataset of Spatio-temporally Localized Atomic Visual Actions
T-CNN	61.3	78.4	76.9	Tube Convolutional Neural Network (T-CNN) for Action Detection in Videos
Two-in-one Two Stream	-	-	74.74	Dance with Flow: Two-in-One Stream Action Detection
DTS	-	76.1	74.3	Finding Action Tubes with a Sparse-to-Dense Framework	-
TACNet	65.5	74.1	73.4	TACNet: Transition-Aware Context Network for Spatio-Temporal Action Detection	-
MR-TS R-CNN	58.5	74.3	73.09	Multi-region two-stream R-CNN for action detection	-
TS R-CNN	56.9	71.1	70.6	Multi-region two-stream R-CNN for action detection	-
Two-in-one	-	-	57.96	Dance with Flow: Two-in-One Stream Action Detection
Actionnness	-	-	56.4	Actionness Estimation Using Hybrid Fully Convolutional Networks	-
Action Tubes	36.2	-	53.3	Finding Action Tubes
Actionness	39.9	-	-	Actionness Estimation Using Hybrid Fully Convolutional Networks	-

0 of 17 row(s) selected.

Action Detection On J Hmdb | SOTA | HyperAI