Image Classification on GTSR (test)
99.7Top-1 AccuracySTDNN
Evaluation results
| Method | Links | |
|---|---|---|
| STDNNNumber of parameters=14M2018.03 | 99.7 | |
| HLSGDNumber of parameters=23.2M2018.03 | 99.6 | |
| MCDNNNumber of parameters=38.5M2018.03 | 99.5 | |
| VGG-B(P)Backbone=VGG-B, Pre-training=pre-trained on ImageNet, #par=82017.05 | 99 | |
| MicronNetNumber of parameters=0.51M, Precision=fp162018.03 | 98.9 | |
| Human2018.03 | 98.8 | |
| CDNNNumber of parameters=1.54M, Evaluation note=average reported top-1 accuracy2018.03 | 98.5 | |
| VGG-B(S)Backbone=VGG-B, Pre-training=trained from scratch, #par=82017.05 | 98.2 | |
| MicronNetNumber of parameters=0.51M, Precision=fixed162018.03 | 98 | |
| DAN_imagenetBackbone=VGG-B, Controller Initialization=ImageNet, #par=2.762017.05 | 97.6 | |
| DAN_imagenet+sketchBackbone=VGG-B, Controller Initialization=ImageNet + Sketch, #par=3.322017.05 | 97.6 | |
| DAN_caltech-256Backbone=VGG-B, Controller Initialization=caltech-256, #par=2.542017.05 | 93.6 | |
| DAN_sketchBackbone=VGG-B, Controller Initialization=sketch, #par=2.542017.05 | 93.3 | |
| DAN_noiseBackbone=VGG-B, Controller Initialization=random weights, #par=1.762017.05 | 90.9 |