Xuan Dong, Weixin Li, Xiaojie Wang, Yunhong Wang.In AAAI,2019
This is the implementation code of AAAI2019's paper "Learning a Deep Convolutional Network for Colorization in Monochrome-Color Dual-Lens System".The example coloring result of gray image is shown in the figure below.
Download the original paper.
Learning a Deep Convolutional Network for Colorization in Monochrome-Color Dual-Lens System.pdf
Clone the repository.
git clone https://github.com/bupt-wx/Deep-Convolutional-Network-For-Colorization-In-Monochrome-Color-Dual-Lens-System.git
Required environment version information.
Tensorflow 1.8.0; Keras 2.1.6; Python 3.6
The algorithm is divided into rough coloring and color correction.
You can test this project by using the following commands and using the images in the Sample_input folder.It should be noted that the algorithm uses the Ycbcr color space, and the pre-processing and post-processing of the algorithm requires converting the color space of the image.
The first step-rough colorization, using test files in RoughColorization folder.The test command is as follows:
python rough_colorization_test.py -fpath test_input_file_path -outpath test_output_file_path
Please replace "test_input_file_path" with the input image path to be tested and "test_output_file_path" with the output image path after testing.
The second step-color correction, using test files in Correction folder.The test command is as follows::
python colorization_correction_test.py -rpath rough_colorization_result_file_path -gpath guided_image_file_path -outpath test_output_file_path
Please replace "rough_colorization_result_file_path" with the results obtained by the first step of rough coloring, "guided_image_file_path" with the input grayscale image and "test_output_file_path" with the output image path after testing.The results should match the images in the Sample_out folder.