This project does colourisation of black-and-white images using an encoder-decoder network with the fusion of a depth estimate.
This was trained on a sample of the COCO dataset for 2 days using google coLab.
Feel free to use the google colab notebook to see the results.
Test images are from the COCO dataset, further training will be done to try improve quality of results
More information and data will be released later
A lot of the code is influenced or taken off https://github.com/CSAILVision/semantic-segmentation-pytorch for the encoder-decoder and an off-the-shelf depth estimate "DiverseDepth" from https://github.com/YvanYin/DiverseDepth