The goal is to use GANs (not neceessarily the DCGAN but others too) to produce funny deep fakes.
I have a data/trump folder which contains Trump's funny faces. The DCGAN creates new fake ones.
I've also uploaded some Alien Crop Circles, which need pre-processing into single images. We can also group them into year or whatever.
Feel free to upload your own dataset of images, provided its not offensive or huge (this is github and there are disk limits).
Have fun, enjoy playing with GANs and read the following hints and tips:
- Pleasse use PEP8 when formating your code
- No need to worry about unit tests, provider your script runs and produces a model
- Please stick to PyTorch, PyTorch-Lightning, TorchMetrics, TorchVision, etc.
- Ensure you have a decent GPU to run PyTorch. Don't try this on the CPU!
- If you get a good model that does something hilarious, feel free to open a PR asking for hosting, and we'll figure something out.
- Please document your Model/Script so others can read and understand what you did and why
Peace.