List your company

Image Classification on ImageNet-1K (val) (Top-1 Accuracy, Throughput, and Latency)

84.3Top-1 AccuracyAS-MLP-B
74.52477.06279.682.138May 31, 2021Jun 8, 2021Jun 16, 2021Jun 24, 2021Jul 2, 2021Jul 10, 2021Jul 18, 2021

Evaluation results

MethodLinks
AS-MLP-BInput Resolution=384 x 384, Params=88M, FLOPs=44.6G, Throughput (image / s)=179.22021.0784.3——
Swin-BInput Resolution=384 x 384, Params=88M, FLOPs=47.0G, Throughput (image / s)=84.72021.0784.2——
EffNet-B6Input Size=528x528, Parameters=43M, FLOPs=19.0G2021.058496.41,317
MSG-Transformer-BInput Size=224x224, Parameters=84M, FLOPs=14.2G2021.0584267.6424
EffNet-B5Input Size=456x456, Parameters=30M, FLOPs=9.9G2021.0583.6168.5768
EfficientNet-B5Input Resolution=456 x 456, Params=30M, FLOPs=9.9G, Throughput (image / s)=169.12021.0783.6——
MSG-Transformer-SInput Size=224x224, Parameters=56M, FLOPs=8.4G2021.0583.4422.5272
Swin-BInput Size=224x224, Parameters=88M, FLOPs=15.4G2021.0583.3257.6547
Swin-BInput Resolution=224 x 224, Params=88M, FLOPs=15.4G, Throughput (image / s)=278.12021.0783.3——
AS-MLP-BInput Resolution=224 x 224, Params=88M, FLOPs=15.2G, Throughput (image / s)=455.22021.0783.3——
ViP-Large/7Input Resolution=224 x 224, Params=88M, Throughput (image / s)=2982021.0783.2——
DeiT-BInput Size=384x384, Parameters=87M, FLOPs=55.4G2021.0583.181.11,213
AS-MLP-SInput Resolution=224 x 224, Params=50M, FLOPs=8.5G, Throughput (image / s)=619.52021.0783.1——
Swin-SInput Size=224x224, Parameters=50M, FLOPs=8.7G2021.0583396.6346
Swin-SInput Resolution=224 x 224, Params=50M, FLOPs=8.7G, Throughput (image / s)=436.92021.0783——
RegY-16GInput Size=224x224, Parameters=84M, FLOPs=16.0G2021.0582.9324.6424
EffNet-B4Input Size=380x380, Parameters=19M, FLOPs=4.2G2021.0582.9345315
RegNetY-16GFInput Resolution=224 x 224, Params=84M, FLOPs=15.9G, Throughput (image / s)=334.72021.0782.9——
TNT-BInput Size=224x224, Parameters=66M, FLOPs=14.1G2021.0582.8231.1414
ViP-Medium/7Input Resolution=224 x 224, Params=55M, Throughput (image / s)=4182021.0782.7——
MSG-Transformer-TInput Size=224x224, Parameters=25M, FLOPs=3.8G2021.0582.4726.5157
PVT-LargeInput Resolution=224 x 224, Params=61M, FLOPs=9.8G2021.0782.3——
T2T-ViTt-24Input Size=224x224, Parameters=64M, FLOPs=13.2G2021.0582.2268.2436
CoaT-Lite-SInput Size=224x224, Parameters=20M, FLOPs=4.0G2021.0581.9——
DeiT-BInput Size=224x224, Parameters=87M, FLOPs=17.5G2021.0581.8278.9393
DeiT-B/16Input Resolution=224 x 224, Params=86M, FLOPs=17.6G, Throughput (image / s)=292.32021.0781.8——
RegY-8GInput Size=224x224, Parameters=39M, FLOPs=8.0G2021.0581.7545.5250
PVT-LargeInput Size=224x224, Parameters=61M, FLOPs=9.8G2021.0581.7337.1338
RegNetY-8GFInput Resolution=224 x 224, Params=39M, FLOPs=8.0G, Throughput (image / s)=591.62021.0781.7——
gMLP-BInput Resolution=224 x 224, Params=73M, FLOPs=15.8G2021.0781.6——
ViP-Small/7Input Resolution=224 x 224, Params=25M, Throughput (image / s)=7192021.0781.5——
T2T-ViTt-19Input Size=224x224, Parameters=39M, FLOPs=8.4G2021.0581.4377.3314
TNT-SInput Size=224x224, Parameters=24M, FLOPs=5.2G2021.0581.3387.1215
Swin-TInput Size=224x224, Parameters=28M, FLOPs=4.5G2021.0581.3692.1189
Swin-TInput Resolution=224 x 224, Params=29M, FLOPs=4.5G, Throughput (image / s)=755.22021.0781.3——
AS-MLP-TInput Resolution=224 x 224, Params=28M, FLOPs=4.4G, Throughput (image / s)=1047.72021.0781.3——
T2T-ViTt-14Input Size=224x224, Parameters=22M, FLOPs=5.2G2021.0580.7559.3225
S2-MLP-deepInput Resolution=224 x 224, Params=51M, FLOPs=9.7G2021.0780.7——
ViP-Small/14Input Resolution=224 x 224, Params=30M, Throughput (image / s)=7892021.0780.5——
RegY-4GInput Size=224x224, Parameters=21M, FLOPs=4.0G2021.0580930.1138
S2-MLP-wideInput Resolution=224 x 224, Params=68M, FLOPs=13.0G2021.0780——
DeiT-SInput Size=224x224, Parameters=22M, FLOPs=4.6G2021.0579.8898.3118
PVT-SmallInput Size=224x224, Parameters=25M, FLOPs=3.8G2021.0579.8749146
ResMLP-36Input Resolution=224 x 224, Params=45M, FLOPs=8.9G, Throughput (image / s)=478.72021.0779.7——
gMLP-SInput Resolution=224 x 224, Params=20M, FLOPs=4.5G2021.0779.4——
ViT-B/16Input Size=384x384, Parameters=87M, FLOPs=55.4G2021.0577.981.11,218
ViT-B/16Input Resolution=384 x 384, Params=86M, FLOPs=55.5G, Throughput (image / s)=85.92021.0777.9——
ViT-L/16Input Size=384x384, Parameters=307M, FLOPs=190.7G2021.0576.526.34,420
Mixer-B/16Input Resolution=224 x 224, Params=59M, FLOPs=11.7G2021.0776.4——
FFInput Resolution=224 x 224, Params=62M, FLOPs=11.4G2021.0774.9——