dilation-tensorflow

A native Tensorflow implementation of semantic segmentation according to Multi-Scale Context Aggregation by Dilated Convolutions by Yu and Koltun.

Pretrained weights have been converted to TensorFlow from the original Caffe implementation.

Currently, only the model pretrained on CityScapes is available. I plan to convert soon also the model trained on CamVid dataset.

You you're looking instead for a Keras+Theano implementation of this very same network you can find it here.

Test image (input)	Test image (prediction)

How-to

1. Download pretrained weights from here:

   CityScapes weights

2. Move weights file into data directory.

3. Run the model on the test image by executing main_tf.py

Configuration

This model has been tested with the following configuration:

• Ubuntu 16.04
• python 3.5.2
• tensorflow 1.1.0
• cv2 3.2.0