MXNet Forum

Mxnet model gallery


#1

I found the following repository of mxnet models: https://github.com/dmlc/mxnet-model-gallery

These models are trained on Imagenet, but the exact training script is not provided. I am particularly interested in what color normalizations were used during training.

Can someone confirm that the same numbers were used as here: https://github.com/apache/incubator-mxnet/blob/master/example/image-classification/train_imagenet.py ?

Thanks


#2

These are very old models. I recommend you stick to the Gluon model zoo. Is there a specific model that is in the model gallery and not in model zoo?


#3

Yes, Inception-BN. It is still used in many research papers as baseline.


#4

We have inception-v3 in gluoncv https://gluon-cv.mxnet.io/model_zoo/classification.html#others


#5

The paper whose results I am trying to reproduce use inception-bn, as far as I understand that is inception-v2.