HyperAI超神经

首页算力平台文档资讯论文教程数据集百科 SOTA LLM 模型天梯 GPU 天梯顶会

中文

HyperAI超神经

Action Recognition In Videos On Ava V21

评估指标

mAP (Val)

评测结果

各个模型在此基准测试上的表现结果

		Paper Title	Repository
STAR/L	41.7	End-to-End Spatio-Temporal Action Localisation with Video Transformers	-
ACAR-Net, SlowFast R-101 (Kinetics-400 pretraining)	30.0	Actor-Context-Actor Relation Network for Spatio-Temporal Action Localization
JMRN + SlowFast-R101-NL	28.4	Pose And Joint-Aware Action Recognition
SlowFast++ (Kinetics-600 pretraining, NL)	28.3	SlowFast Networks for Video Recognition
LFB (Kinetics-400 pretraining)	27.7	Long-Term Feature Banks for Detailed Video Understanding
I3D Tx HighRes	27.6	Video Action Transformer Network	-
SlowFast (Kinetics-600 pretraining, NL)	27.3	SlowFast Networks for Video Recognition
SlowFast (Kinetics-600 pretraining)	26.8	SlowFast Networks for Video Recognition
SlowFast (Kinetics-400 pretraining)	26.3	SlowFast Networks for Video Recognition
I3D I3D	23.4	Video Action Transformer Network	-
D3D (ResNet RPN, Kinetics-400 pretraining)	23	D3D: Distilled 3D Networks for Video Action Recognition
I3D w/ RPN + JFT (Kinetics-400 pretraining(	22.8	A Better Baseline for AVA	-
S3D-G w/ ResNet RPN (Kinetics-400 pretraining(	22.0	AVA: A Video Dataset of Spatio-temporally Localized Atomic Visual Actions
I3D w/ RPN (Kinetics-400 pretraining(	21.9	A Better Baseline for AVA	-
ARCN	17.4	Actor-Centric Relation Network

0 of 15 row(s) selected.

Action Recognition In Videos On Ava V21 | SOTA | HyperAI超神经