search

isht7 / pytorch-deeplab-resnet

DeepLab resnet v2 model in pytorch
MIT License
602 stars 118 forks source link

Why isn't the input to model normalized? #11

Closed rohitgajawada closed 7 years ago

rohitgajawada commented 7 years ago

Hi! I observed that your input to your model is not normalized between 0 and 1 and even has values in the 140s. Is there any reason why you did this?

isht7 commented 7 years ago

This is not an issue. Please do not ask such things here. That is how images are typically passed in CNNs. This is an implementation of project written in caffe, and that is how caffe does it. Other(than images) types of data, they are generally normalized.

isht7 commented 7 years ago

you can email me directly for questions that are not issues at isht dot dwivedi at gmail dot com

isht7 commented 7 years ago

To be clear: Image is read(image values 0-255), mean is subtracted and the image is passed as it is(after augmentation) to the CNN.