List your company
ResearchFormerly WizWand

Action Recognition on HMDB51 (all splits)

80.7AccuracyTwo-Stream I3D
50.43658.29366.1574.007Dec 3, 2018Jan 4, 2019Feb 5, 2019Mar 9, 2019Apr 10, 2019May 12, 2019Jun 13, 2019
Updated 8mo ago

Evaluation results

MethodLinks
Two-Stream I3DPre-train dataset=ImageNet & Kinetics, Input frames=64f2018.1280.7
LGD-3D Two-streamPretraining=ImageNet+Kinetics-6002019.0680.5
I3D Two-streamPretraining=ImageNet+Kinetics-4002019.0680.2
LGD-3D FlowPretraining=ImageNet+Kinetics-6002019.0678.9
Two-Stream R(2+1)DPre-train dataset=Kinetics2018.1278.7
R(2+1)D Two-streamPretraining=Kinetics-4002019.0678.7
R(2+1)D FlowPretraining=Kinetics-4002019.0676.4
S3D-G RGBPretraining=ImageNet+Kinetics-4002019.0675.9
LGD-3D RGBPretraining=ImageNet+Kinetics-6002019.0675.7
I3D FlowPretraining=ImageNet+Kinetics-4002019.0674.6
R(2+1)D-RGBPre-train dataset=Kinetics2018.1274.5
I3D RGBPretraining=ImageNet+Kinetics-4002019.0674.5
R(2+1)D RGBPretraining=Kinetics-4002019.0674.5
ResNeXt-101 RGBPretraining=Kinetics-4002019.0670.2
TSNPretraining=ImageNet2019.0669.4
Two-Stream STM-NetsPre-train dataset=ImageNet2018.1268.9
Two-Stream TSNPre-train dataset=ImageNet2018.1268.5
ARTNetPre-train dataset=Kinetics2018.1267.6
RGB I3DPre-train dataset=ImageNet & miniKinetics, Input frames=64f2018.1266.4
3D ResNeXt-101Pre-train dataset=Kinetics2018.1263.8
SUSiNetPre-train dataset=Kinetics, Number of tasks=multi2018.1262.7
IDTPretraining=N/A2019.0661.7
3D ResNet-50Pre-train dataset=Kinetics2018.1261
SUSiNetPre-train dataset=Kinetics, Number of tasks=1-task2018.1260.2
Two-streamPretraining=ImageNet2019.0659.4
FSTCNPre-train dataset=ImageNet2018.1259.1
3D ResNet-18Pre-train dataset=Kinetics2018.1256.4
C3DPre-train dataset=Sports-1M2018.1251.6