Open myzhuang opened 1 year ago
Hello, thanks for your attention to our work. We reviewed the codes and found some problems with the result and the newest results are updated now. The batch-size is 128, learning-rate is 0.001 for Adam, input image size is just 360*240.
Thanks for your share!And can you provide more detail about training setting of MobileNetV3, such as train epoch, learning rate, and input image resloution?