Not getting a good acc
Hi @shicai , I tried your depoly prototxt, yet didn't get a good acc result.
I'd be appreciated if you could share your solver prototxt for me to check.
thanks!
what is your result? solver prototxt file has no effect on the val accuracy of the pretrained model.
Hi @shicai I only got 54 as top 1 acc and 79 as top 5 acc. Btw I didn't use the pretrained model, I'm trying to train the net from the imagenet dataset. thanks!
training models on ImageNet is somehow very tricky. you'd better train some simple models first.
@Shicai First of all Thanks alot. It would be very helpful if you could share the training/testing script as well for replicating the exact results.
@shicai Hi shicai,
I've tuned the top 1 acc to 67%, still sort of a long way from the paper acc. I trained with batch size 256, 90 epochs and decay the lr rate by 0.1 every 30 epochs.
I'd be appreciated if there're any tricks that you could share with me to reach the 70 acc?
thanks a lot
please see: https://github.com/shicai/MobileNet-Caffe/issues/1