search

DrSleep / tensorflow-deeplab-resnet

DeepLab-ResNet rebuilt in TensorFlow
MIT License
1.25k stars 429 forks source link

How does the label encoding work? #99

Closed akshittyagi closed 7 years ago

akshittyagi commented 7 years ago

The label is initially an image read as a 3 channel image with pixel values. How is it converted into a pixel-wise map of labels? As in when does, it get the marking of labels?

Init label image: at each pixel, we have a (3,) vector with channel colors Training label: at each pixel, we have a marking of which class it belongs to

When does this transformation take place?

DrSleep commented 7 years ago

Please read this comment; and check out the official DeepLab FAQ.

The labels are converted into the ones without a colour map before the training process starts.