Pytorch-Colorful Image Colorization

Introduction

An unofficial pytorch implementation of "Colorful image colorization"

I am running the paddlepaddle version of the model and the relevant paddle model weights file will be available here:

I don't have the free time and GPU to run this code these days... So I'm afraid I can't provide model weight files and performance reports for now.

Anyway, if you encounter any problems during the training process, feel free to ask!

ImageNet Dataset I resized all images to 256*256

Training set：1281167 images
Validation set：10000 images that from the imagenet validation set
Test set：10000 images that from the imagenet validation set. Proposed by Learning Representations for Automatic Colorization

# clone this repo
git clone https://github.com/nku-shengzheliu/Pytorch-Colorful-Image-Colorization.git
cd Pytorch-Colorful-Image-Colorization

pip install -r requirements.txt

Train the model using the following commands:

python train.py

If training is interrupted, you can resume it with the --resume parameter, which sets --resume to the last saved weight file.

python demo.py --data_path {data path} --split {train/val/test}  --model_path {saved_path/XXXX.pth} --save_path{path to save visualized results}