Action Classification On Charades

Metrics

MAP

Results

Performance results of various models on this benchmark

Model Name	MAP	Paper Title	Repository
AssembleNet++ 50	59.8	AssembleNet++: Assembling Modality Representations via Attention Connections
SlowFast (Kinetics-600 pretraining, NL)	45.2	SlowFast Networks for Video Recognition
TokenLearner	66.3	TokenLearner: What Can 8 Learned Tokens Do for Images and Videos?
Asyn-TF	22.4	Asynchronous Temporal Fields for Action Recognition
JMRN (Pose only)	16.2	Pose And Joint-Aware Action Recognition
PoTion + (GCN + I3D + NL I3D)	40.8	PoTion: Pose MoTion Representation for Action Recognition	-
VicTR (ViT-L/14)	57.6	VicTR: Video-conditioned Text Representations for Activity Recognition	-
STLT + I3D	38.5	Revisiting spatio-temporal layouts for compositional action recognition
LFB	42.5	Long-Term Feature Banks for Detailed Video Understanding
AdaFocus (weak supervision, MViT-B-K400-pretrain, 16x4)	41.4	Towards Weakly Supervised End-to-end Learning for Long-video Action Recognition	-
BIKE	50.7	Bidirectional Cross-Modal Knowledge Exploration for Video Recognition with Pre-trained Vision-Language Models
DEEP-HAL with ODF+SDF (I3D)	50.16	Self-supervising Action Recognition by Statistical Moment and Subspace Descriptors	-
MoViNet-A2	32.5	MoViNets: Mobile Video Networks for Efficient Video Recognition
ActionCLIP (ViT-B/16)	44.3	ActionCLIP: A New Paradigm for Video Action Recognition
MViT-B, 32x3 (Kinetics-400 pretraining)	44.3	Multiscale Vision Transformers
JMRN + R101-NL-LFB	43.23	Pose And Joint-Aware Action Recognition
TubeViT-L	66.2	Rethinking Video ViTs: Sparse Video Tubes for Joint Image and Video Learning
MViT-B-24, 32x3 (Kinetics-600 pretraining)	47.7	Multiscale Vision Transformers
HAF+BoW/FV/OFF halluc. +MSK×8/PN	43.1	Hallucinating IDT Descriptors and I3D Optical Flow Features for Action Recognition with CNNs	-
I3D	32.9	Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset

0 of 49 row(s) selected.