[CADL'22, ECCVW] Official repository of paper titled "EdgeNeXt: Efficiently Amalgamated CNN-Transformer Architecture for Mobile Vision Applications".
EdgeNeXt-B model ImageNet-21K weights are released. The model is first pretrained on ImageNet-21K followed by fine tuning on ImageNet-1K. We follow the ConvNeXt training recipe for ImageNet-21K pretraining. Our EdgeNeXt-B ImageNet-21K pretrained model achieves 83.3% top-1 ImageNet-1K accuracy.
EdgeNeXt-B model and pretrained weights are released. It achieves 82.5% top-1 ImageNet-1K accuracy with 18.51M parameters and 3.84G MAdds. Further, using USI (https://arxiv.org/abs/2204.03475) training recipe, the same model achieves 83.7% accuracy.
EdgeNeXt trained using USI (https://arxiv.org/abs/2204.03475). It achieves 81.1% top-1 ImageNet-1K accuracy.
Release of evaluation and training code of EdgeNeXt classification models. All the pretrained weights are also provided.