Work to create a GAN for generating novel pottery designs.
In progress - data collection process mostly complete. Dataset cleanup and GAN architecture fine tuning to be confirmed. First pass using Wasserstein GP GAN architecture produces promising results, but poor fine details in generated images.
-
To run image scraping/processing scripts - follow the steps on Alexey B's fork of darknet: https://github.com/AlexeyAB/darknet#how-to-use-yolo-as-dll-and-so-libraries. Note that installing Visual Studio first, then CUDA+CUDNN is required to get it to build with CUDA support. Build darknet.exe, and yolo_cpp_dll.dll. (https://github.com/AlexeyAB/darknet#how-to-use-yolo-as-dll-and-so-libraries )
- Place pthreadVC2.dll, yolo_cpp_dll.dll in scripts folder.
- Place yolov4.weights in scripts/cfg folder (links to download in Darknet repo)