BunnyShan

@zymazloa mobilenet need to use NHWC, and if you use synthetic_gpu_images, the data_format must be set to NHWC. Also I notice mobilenet_v2 training speed is about 250images/sec with the right...

@zhreshold Another question: Why SSD's input need to be resized to a predefine shape? SSD is a fcn framework, will no resize operation increases the performance?

I see. In yolo2, it uses the same size in one batch and change another size after a few batches. I will look for some other solutions. Thank you!

@zhreshold I got about mAP:65% on voc after update the new code. Is there something wrong?

Thanks a lot. Fix gt_count will cause recall decrease. I will test it too.

The gt_count fix only cause about 1% mAP drop.

@zhreshold Is there any progress?

@zhreshold why update rescale_grad like this `'rescale_grad': 1.0 / len(ctx) if len(ctx) > 0 else 1.0` ? And previous `'rescale_grad': 1.0` makes my model don't converge.

hi, @zhreshold is resnet50 broken? I can't reproduce the result either. Could you plz give me some suggestion?

@zhreshold How did you get this? Could you please share your configuration?