#Colorization By nilboy

A Tensorflow implementation of ECCV2016 paper(Colorful Image Colorization)

###Train

1. Download imagenet data and Extract to one directory named 'Imagenet'

2. link the Imagenet data directory to this project path

   ln -s $Imagenet data/imagenet
   

python tools/create_imagenet_list.py

python tools/train.py -c conf/train.cfg

1. transform your training data to text_record file

2. calculate the training data prior-probs(reference to tools/create_prior_probs.py)

3. write your own train-configure file

4. train (python tools/train.py -c $your_configure_file)

###test demo

1. Download pretrained model(https://drive.google.com/file/d/0B-yiAeTLLamRWVVDQ1VmZ3BxWG8/view?usp=sharing)

   mv color_model.ckpt models/model.ckpt
   

2. Test

   python demo.py
   

''' ssh xieya@biwirender15 ". /srv/glusterfs/xieya/anaconda2/etc/profile.d/conda.sh; export PATH="/srv/glusterfs/xieya/cuda-9.0/bin${PATH:+:${PATH}}"; export LD_LIBRARY_PATH="/srv/glusterfs/xieya/cuda-9.0/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}"; echo $PATH; conda activate; cd ~/colorization-gan-7; CUDA_VISIBLE_DEVICES=5 python -u tools/train.py -c conf/train.cfg > ~/gan7.log" '''