Based on pix2pix-tensorflow which is tensorflow port of pix2pix by Isola et al. Visit raihan-tanvir.github.io/generative-jamdani to see the live demonstration.
Sample output generated by our Generative-Jamdani from input sketches, like these examples from the original paper:
This is an implementation of pix2pix-tensorflow on Jamdani Noksha dataset. It is meant to be a faithful implementation of the original work and so does not add anything.
- Tensorflow 1.4.1
- Linux with Tensorflow GPU edition + cuDNN
# clone this repo
git clone https://github.com/raihan-tanvir/generative-jamdani.git
cd generative-jamdani
python pix2pix.py \
--mode train \
--output_dir model/ \
--max_epochs 150 \
--input_dir dataset/train \
--which_direction BtoA
# test the model
python pix2pix.py \
--mode test \
--output_dir output/ \
--input_dir dataset/test \
--checkpoint model/The test run will output an HTML file at output/index.html that shows input/output/target image sets.
The data format used by this program is the same as the original pix2pix format, which consists of images of input and desired output side by side like:
For example:
| dataset | example |
|---|---|
| JamdaniNoksha-B Dataset 1116 images |
 |
| JamdaniNoksha-BRes Dataset 1983 images |
 |
| JamdaniNoksha-Skel Dataset 7932 images |
 |
| JamdaniNoksha-RedB Dataset 913 images |
 |
| JamdaniNoksha-Sketch Dataset 910 images |
 |
@inproceedings{generative-jamdani,
title={Jamdani Motif Generation using Conditional GAN},
author={Tanvir Rouf Shawon, Raihan Tanvir, Humaira Ferdous Shifa, Susmoy Kar, Mohammad Imrul Jubair},
publisher={International Conference on Computer and Information Technology},
venue={Ahsanullah University of Science & Technology, Dhaka, Bangadesh},
year={2020}
}
This is a implementation of pix2pix-tensorflow on Jamdani Noksha dataset. Thanks to the Tensorflow Affinelayer for making such a wonderful port! And special thanks to Phillip Isola for the original pix2pix.