Action Recognition In Videos On Hmdb 51

评估指标

Average accuracy of 3 splits

评测结果

各个模型在此基准测试上的表现结果

		Paper Title	Repository
VideoMAE V2-g	88.1	VideoMAE V2: Scaling Video Masked Autoencoders with Dual Masking
DEEP-HAL with ODF+SDF(I3D)	87.56	Self-supervising Action Recognition by Statistical Moment and Subspace Descriptors	-
TO+MaxExp+IDT	87.21	High-order Tensor Pooling with Attention for Action Recognition	-
SCK⊕(I3D)+IDT	86.11	Tensor Representations for Action Recognition
SO+MaxExp+IDT	85.70	High-order Tensor Pooling with Attention for Action Recognition	-
R2+1D-BERT	85.10	Late Temporal Modeling in 3D CNN Architectures with BERT for Action Recognition
Ours + ResNext101 BERT	84.53	Pose And Joint-Aware Action Recognition
SMART	84.36	SMART Frame Selection for Action Recognition	-
OmniSource (SlowOnly-8x8-R101-RGB + I3D Flow)	83.8	Omni-sourced Webly-supervised Learning for Video Recognition
ZeroI2V ViT-L/14	83.4	ZeroI2V: Zero-Cost Adaptation of Pre-trained Transformers from Image to Video
PERF-Net (distilled S3D-G)	83.2	PERF-Net: Pose Empowered RGB-Flow Net	-
BIKE	83.1	Bidirectional Cross-Modal Knowledge Exploration for Video Recognition with Pre-trained Vision-Language Models
BubbleNET	82.60	Bubblenet: A Disperse Recurrent Structure To Recognize Activities	-
HAF+BoW/FV halluc	82.48	Hallucinating IDT Descriptors and I3D Optical Flow Features for Action Recognition with CNNs	-
CCS + TSN (ImageNet+Kinetics pretrained)	81.9	Cooperative Cross-Stream Network for Discriminative Action Representation	-
RepFlow-50 ([2+1]D CNN, FcF, Non-local block)	81.1	Representation Flow for Action Recognition
Multi-stream I3D	80.92	Contextual Action Cues from Camera Sensor for Multi-Stream Action Recognition	-
Two-stream I3D	80.9	Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset
MARS+RGB+FLow (64 frames, Kinetics pretrained)	80.9	MARS: Motion-Augmented RGB Stream for Action Recognition	-
Two-Stream I3D (Imagenet+Kinetics pre-training)	80.7	Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset

0 of 76 row(s) selected.