SiamFC-TensorFlow icon indicating copy to clipboard operation
SiamFC-TensorFlow copied to clipboard

关于data augmentation

Open tongtybj opened this issue 6 years ago • 2 comments

请问这里的transformation为什么要 -8 ? https://github.com/bilylee/SiamFC-TensorFlow/blob/master/datasets/dataloader.py#L38

另外,因为下面的处理,instance embeds的shape永远都是[batch_size, 255-28, 255-28, 3]吗? https://github.com/bilylee/SiamFC-TensorFlow/blob/master/datasets/dataloader.py#L39

tongtybj avatar Mar 18 '20 16:03 tongtybj

Is it because the random stretch may resize the raw image with a -0.05: https://github.com/bilylee/SiamFC-TensorFlow/blob/f572dca95f2b3b2861f54de467259753428e468c/datasets/transforms.py#L49?

tongtybj avatar Mar 18 '20 18:03 tongtybj

这是为了对训练图像做 +- 4 个 pixel 的平移。如果直接 RandomCrop(255 - 2 * 8) 会导致平移 +- 8 个 pixel,所以先CenterCrop,再 RandomCrop。相关问题 #19 #32 #56

bilylee avatar Jun 04 '20 01:06 bilylee