search

aurora95 / Keras-FCN

Keras-tensorflow implementation of Fully Convolutional Networks for Semantic Segmentation（Unfinished）
MIT License
650 stars 268 forks source link

regarding adopting the code to fit my own dataset #20

Open wenouyang opened 7 years ago

wenouyang commented 7 years ago

Hi aurora95,

Thanks for sharing the code.

I am planning to adopt your code to training my own data set. The training images are of size 500*700, and each of which is black-and-white, i.e., with only one single channel. What kind of modifications have to be made to make your code fitful for my data set? Thanks a lot.

Wenouyang

ahundt commented 7 years ago

The changes wouldn't be too difficult, just try it and add minor fixes when 3 channels are expected but there is only one. If you don't want to spend time on those bugfixes the shortcut would be to simply repeat your 1 channel image 3 times, if RGB are all equal then you've still got a grayscale image.

ahundt commented 7 years ago

Oh and the pretrained weight transfer from imagenet would likely not "just work" on a 1 channel image.