List your company

Image Classification on ImageNet-1K (val) (Hardware Latency Profile)

80.1Top-1 AccMCTF
66.89270.32173.7577.179Jun 8, 2022Sep 23, 2022Jan 9, 2023Apr 27, 2023Aug 12, 2023Nov 28, 2023Mar 15, 2024

Evaluation results

MethodLinks
MCTFBackbone=DeiT-S, FLOPs (G)=2.6, Params (M)=22, r=162024.0380.1————
DeiT-SBackbone=DeiT-S, FLOPs (G)=4.6, Params (M)=222024.0379.8————
ATSBackbone=DeiT-S, FLOPs (G)=2.9, Params (M)=222024.0379.7————
BATBackbone=DeiT-S, FLOPs (G)=3.0, Params (M)=222024.0379.6————
EVITBackbone=DeiT-S, FLOPs (G)=3.0, Params (M)=222024.0379.5————
MobileOne-S4Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0679.426.61.860.952.17
EvoViTBackbone=DeiT-S, FLOPs (G)=3.0, Params (M)=222024.0379.4————
ToMeBackbone=DeiT-S, FLOPs (G)=2.7, Params (M)=222024.0379.4————
Dynamic ViTBackbone=DeiT-S, FLOPs (G)=2.9, Params (M)=232024.0379.3————
SPVITBackbone=DeiT-S, FLOPs (G)=2.6, Params (M)=222024.0379.3————
IA-RED²Backbone=DeiT-S, FLOPs (G)=3.2, Params (M)=222024.0379.1————
RepVGG-B2Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0678.8492.86.384.796.83
A-ViTBackbone=DeiT-S, FLOPs (G)=3.6, Params (M)=222024.0378.6————
RepVGG-B1Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0678.4193.73.733.174.28
MobileOne-S3Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0678.116.471.530.761.28
MobileOne-S2Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0677.414.871.180.721.07
EfficientNet-B0Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0677.128.711.721.352.49
RepVGG-A2Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0676.593.432.412.412.28
MobileOne-S1Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0675.913.040.890.660.79
MNASNet-A1Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0675.824.0610.950.88
MobileNetV3-LPrecision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU), H-swish replacement=ReLU2022.0675.217.091.093.81.01
RepVGG-B0Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0675.155.971.821.421.43
MobileNetV2-x1.4Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0674.715.671.360.80.98
RepVGG-A1Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0674.547.151.681.421.21
MCTFBackbone=DeiT-T, FLOPs (G)=0.7, Params (M)=5, r=162024.0372.7————
RepVGG-A0Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0672.443.611.231.281.01
BATBackbone=DeiT-T, FLOPs (G)=0.8, Params (M)=52024.0372.3————
DeiT-TBackbone=DeiT-T, FLOPs (G)=1.2, Params (M)=52024.0372.2————
SPVITBackbone=DeiT-T, FLOPs (G)=0.9, Params (M)=52024.0372.1————
MobileNetV2-x1.0Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.067213.650.980.690.77
EvoViTBackbone=DeiT-T, FLOPs (G)=0.8, Params (M)=52024.0372————
MobileOne-SOPrecision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0671.410.550.790.560.59
ToMeBackbone=DeiT-T, FLOPs (G)=0.7, Params (M)=52024.0371.3————
A-ViTBackbone=DeiT-T, FLOPs (G)=0.8, Params (M)=52024.0371————
MobileNet V1Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0670.610.650.950.580.73
ShuffleNetV2-x1.0Precision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU)2022.0669.416.60.684.58—
MobileNetV3-SPrecision=fp16 (except Pixel-6+ TPU), int8 (Pixel-6+ TPU), H-swish replacement=ReLU2022.0667.410.380.833.740.67