List your company
ResearchFormerly WizWand

Action Recognition on UCF-101

99.7Top-1 AccFTP-UniFormerV2-L/14
84.51688.45892.496.342Aug 9, 2020Jul 27, 2021Jul 14, 2022Jul 1, 2023Jun 17, 2024Jun 4, 2025May 22, 2026
Updated 5mo ago

Evaluation results

MethodLinks
FTP-UniFormerV2-L/14Params=402M2024.0399.7——
Video-STAR-7BProtocol=Base-to-Novel (harmonic mean)2026.0399.7——
VideoMAE V2Params=1050M2024.0399.6——
Video-STAR-7BSetting=Full validation set, Protocol=Cross-dataset2025.1099.4——
Video-STAR-7BSetting=Three validation splits, Protocol=Cross-dataset2025.1099.4——
IMD-BPretrained Weights=Kinetics2022.0399.1——
MM-ViTPretrained Weights=Kinetics2022.0398.9——
BIKE (ViT-L)Pre-training=Kinetics, Modality=RGB, Backbone=ViT-L2022.1298.8——
FTP-UniFormerV2-B/16Params=136M2024.0398.7——
R(2+1)D-BERTInput frames=64, Single GFLOPs=152.97, MParams=66.672021.0898.69——
BIKE (ViT-L, 336)Pre-training=Kinetics, Modality=RGB, Backbone=ViT-L, Resolution=3362022.1298.6——
UniFormerV2-L/14Params=354M2024.0398.2——
Two-stream I3DBackbone=3D Inception-v1, Pre-train=Kinetics, #3D=true, #Optical flow=true, #Frames=64+642021.0397.8——
MVD-LParams=87M2024.0397.5——
TDNPre-training=Kinetics, Modality=RGB2022.1297.4——
InternVideo2-s2-6BSetting=16 x 224, Protocol=Linear probing2024.0397.3——
InternVideo2Protocol=Fine-Tuned2026.0397.3——
LGD-3D RGBInput frames=16, Views=152021.0897——
R(2+1)DPre-training=Kinetics, Modality=RGB2022.1296.8——
S3D-GPre-training=Kinetics, Modality=RGB2022.1296.8——
UniFormerV2-B/16Params=115M2024.0396.8——
Video-STAR-3BSetting=Full validation set, Protocol=Cross-dataset2025.1096.8——
Video-STAR-3BSetting=Three validation splits, Protocol=Cross-dataset2025.1096.7——
MVFNetPre-training=Kinetics, Modality=RGB2022.1296.6——
Ours-cumulativeBackbone=TSM-ResNet50, Pre-train=Kinetics, #Frames=all, #Clusters=82021.0396.4——
InternVideo2-s1-6BSetting=16 x 224, Protocol=Linear probing2024.0396.3——
STMBackbone=ResNet50, Pre-train=ImageNet+Kinetics, #Frames=162021.0396.2——
Ours-slopeBackbone=TSM-ResNet50, Pre-train=Kinetics, #Frames=all, #Clusters=82021.0396.2——
STMPre-training=Kinetics, Modality=RGB2022.1296.2——
STMInput frames=16, Views=10, Single GFLOPs=66.5, MParams=22.42021.0896.2——
VideoMAEParams=87M2024.0396.1——
InternVideo2-s1-1BSetting=16 x 224, Protocol=Linear probing2024.0396——
TSMBackbone=ResNet50, Pre-train=Kinetics, #Frames=82021.0395.9——
IMD-BPretrained Weights=IN-21K2022.0395.9——
TSMPre-training=Kinetics, Modality=RGB2022.1295.9——
I3DPre-training=Kinetics, Modality=RGB2022.1295.6——
MM-ViTPretrained Weights=IN-21K2022.0395.4——
I3DPretrained Weights=Kinetics2022.0395.4——
GDTParams=33M2024.0395.2——
MMV FACParams=94M2024.0395.2——
RGB I3DBackbone=3D Inception-v1, Pre-train=Kinetics, #3D=true, #Frames=642021.0395.1——
SI+DI+OF+DOFBackbone=ResNeXt50, Pre-train=Imagenet, #Optical flow=dynamic images2021.0395——
CoVIAR + OFPretrained Weights=IN-1K2022.0394.9——
LIGARInput frames=16, Views=10, Single GFLOPs=4.74, MParams=4.472021.0894.8599.598.61
ECO-En LiteBackbone=BNIncep+3D Res18, Pre-train=Kinetics, #3D=true, #Frames=82021.0394.8——
CVRLParams=328M2024.0394.4——
BDC-CLIPK (Shots)=16, Few-shot setting=true, Fine-tuned from CLIP=true2026.0594.4——
SimVAK (Shots)=16, Few-shot setting=true, Fine-tuned from CLIP=true2026.0594.4——
ARTNetPre-training=Kinetics, Modality=RGB2022.1294.3——
TC-CLIPK (Shots)=16, Few-shot setting=true, Fine-tuned from CLIP=true2026.0594.3——
TSNPretrained Weights=IN-1K2022.0394.2——
XDCParams=33M2024.0394.2——
OSTK (Shots)=16, Few-shot setting=true, Fine-tuned from CLIP=true2026.0593.9——
LIGARInput frames=16, Views=1, Single GFLOPs=4.74, MParams=4.472021.0893.6399.197.82
CORPParams=32M2024.0393.5——
VIMPACParams=307M2024.0392.7——
ViFi-CLIPK (Shots)=16, Few-shot setting=true, Fine-tuned from CLIP=true2026.0592.7——
Two-stream fusionPretrained Weights=IN-1K2022.0392.5——
SimVAK (Shots)=8, Few-shot setting=true, Fine-tuned from CLIP=true2026.0592.4——
TC-CLIPK (Shots)=8, Few-shot setting=true, Fine-tuned from CLIP=true2026.0592——
BDC-CLIPK (Shots)=8, Few-shot setting=true, Fine-tuned from CLIP=true2026.0592——
TVTSv2-HSetting=12 x 224, Protocol=Linear probing2024.0391.8——
TSNBackbone=ResNet50, Pre-train=Kinetics, #Frames=82021.0391.7——
OSTK (Shots)=8, Few-shot setting=true, Fine-tuned from CLIP=true2026.0591.7——
ActionCLIPK (Shots)=16, Few-shot setting=true, Fine-tuned from CLIP=true2026.0591.4——
X-CLIPK (Shots)=16, Few-shot setting=true, Fine-tuned from CLIP=true2026.0591.4——
TSMBackbone=ResNet50, Pre-train=Kinetics, Implementation=our impl., #Frames=12021.0391.2——
DINOv2-gSetting=8 x 224, Protocol=Linear probing2024.0391.2——
OpenCLIP-GSetting=8 x 224, Protocol=Linear probing2024.0390.7——
SimVAK (Shots)=4, Few-shot setting=true, Fine-tuned from CLIP=true2026.0590.5——
CoVIARPretrained Weights=IN-1K2022.0390.4——
TC-CLIPK (Shots)=4, Few-shot setting=true, Fine-tuned from CLIP=true2026.0590.1——
ViFi-CLIPK (Shots)=8, Few-shot setting=true, Fine-tuned from CLIP=true2026.0590——
A5K (Shots)=16, Few-shot setting=true, Fine-tuned from CLIP=true2026.0589.9——
InternVideo2Protocol=Zero-Shot2026.0389.5——
BDC-CLIPK (Shots)=4, Few-shot setting=true, Fine-tuned from CLIP=true2026.0589.3——
Vi2CLRParams=9M2024.0389.1——
CMPLArchitecture=3D-ResNet50, #Frames=82021.1288.9——
CtPPre-training Dataset=K400, Architecture=R(2+1)D2021.0588.4——
SeCoEvaluation Protocol=Fine-Tuning, Pre-train data (duration)=K400 (28d), Modality=V2020.0888.3——
X-CLIPK (Shots)=8, Few-shot setting=true, Fine-tuned from CLIP=true2026.0588.3——
CoCLRPre-training Dataset=K400, Architecture=S3D2021.0587.9——
CoCLRArchitecture=S3D-G, #Frames=322021.1287.6——
OSTK (Shots)=4, Few-shot setting=true, Fine-tuned from CLIP=true2026.0587.5——
PacePredictionArchitecture=S3D-G, #Frames=642021.1287.1——
SimVAK (Shots)=2, Few-shot setting=true, Fine-tuned from CLIP=true2026.0587——
TC-CLIPK (Shots)=2, Few-shot setting=true, Fine-tuned from CLIP=true2026.0586.8——
CtPPre-training Dataset=K400, Architecture=R3D2021.0586.2——
CtPPre-training Dataset=UCF, Architecture=R(2+1)D2021.0586.2——
ImageNet InitArchitecture=3D-ResNet50, #Frames=82021.1286.2——
SimVAWSE=false, Backbone Status=fine-tuned, Setting=Zero-shot, Pre-train=Kinetics-4002026.0586.2——
MemDPCEvaluation Protocol=Fine-Tuning, Pre-train data (duration)=K400 (28d), Modality=VF2020.0886.1——
MemDPCParams=32M2024.0386.1——
SimVAWSE=true, Backbone Status=fine-tuned, Setting=Zero-shot, Pre-train=Kinetics-4002026.0586——
BDC-CLIPWSE=true, Backbone Status=fine-tuned, Setting=Zero-shot, Pre-train=Kinetics-4002026.0585.9——
A5K (Shots)=8, Few-shot setting=true, Fine-tuned from CLIP=true2026.0585.7——
BDC-CLIPK (Shots)=2, Few-shot setting=true, Fine-tuned from CLIP=true2026.0585.6——
TC-CLIPWSE=true, Backbone Status=fine-tuned, Setting=Zero-shot, Pre-train=Kinetics-4002026.0585.4——
STDDSetting=Three validation splits, Protocol=Cross-dataset2025.1085.2——
ViFi-CLIPK (Shots)=4, Few-shot setting=true, Fine-tuned from CLIP=true2026.0585.1——