Open jonbakerfish opened 8 years ago
I noticed that there are different image pre-processing pipelines for different models.
For Inception-BN, the images should be normalized with the mean_224.nd:
mean_224.nd
mean_img = mx.nd.load("Inception/mean_224.nd")["mean_img"] normed_img = sample - mean_img.asnumpy()
For Inception-V3, the images should be:
normed_img = sample - 128. normed_img /= 128.
For Inception-BN-21k, the images should be:
normed_img = sample - 117.
Am I correct with the above settings?
I noticed that there are different image pre-processing pipelines for different models.
For Inception-BN, the images should be normalized with the
mean_224.nd:For Inception-V3, the images should be:
For Inception-BN-21k, the images should be:
Am I correct with the above settings?