Using fully convolutional networks for semantic segmentation with caffe for the cityscapes dataset
Branch: master
Clone or download
Latest commit 9f88d5d Jan 5, 2018
Type Name Latest commit message Commit time
Failed to load latest commit information.
preparation Create Oct 11, 2016
train Update solver.prototxt Oct 11, 2016 Update Jan 4, 2018
final_model_url.txt Update final_model_url.txt Oct 11, 2016
report.pdf Add report Nov 15, 2017
vgg_16_layers_conv_url.txt Update and rename vgg_16_layers_conv_url to vgg_16_layers_conv_url.txt Oct 11, 2016

Using fully convolutional networks for semantic segmentation (Shelhamer et al.) with caffe for the cityscapes dataset

How to get started

  • Download the cityscapes dataset and the vgg-16-layer net
  • Modify the images in the dataset with or for less resource demanding training and evaluation
  • Create the 32 pixel stride net with
  • Modify the paths in train.txt and val.txt (first line: path to training/validation images, second line: path to annotations)
  • Start training with
  • Run to evaluate your model or to create images with pixel label ids


Fully Convolutional Models for Semantic Segmentation:

Shelhamer, Evan, Jonathon Long, and Trevor Darrell. "Fully Convolutional Networks for Semantic Segmentation." PAMI, 2016, URL

Cityscapes Dataset (Semantic Understanding of Urban Street Scenes):

Cordts, Marius, et al. "The cityscapes dataset." CVPR Workshop on The Future of Datasets in Vision. 2015, URL

Caffe Deep Learning Framework:

Jia, Yangqing, et al. "Caffe: Convolutional architecture for fast feature embedding." Proceedings of the 22nd ACM international conference on Multimedia. ACM, 2014, URL