tanminggang/smart-sketch

🖌 photorealistic drawings from simple sketches using NVIDIA's GauGAN

SmartSketch

Supercharge your creativity with state of the art image synthesis

Video Demo below!

Credits

• See project page here: https://nvlabs.github.io/SPADE/
• Read paper here: https://arxiv.org/abs/1903.07291
• See source code here: https://github.com/nvlabs/spade/
• Special thanks to @AndroidKitKat for helping us host this!

Set Up

• You'll need to install the pretrained generator model for the COCO dataset into checkpoints/coco_pretrained/. Instructions for this can be found on the nvlabs/spade repo.

• Make sure you install all the Python requirements using pip3 install -r requirements.txt (in /backend folder).

• Once you do so, you should be able to run the server using python3 server.py. It will run it on 0.0.0.0 on port 80 (on 127.0.0.1 for Windows users). Unfortunately, these are hardcoded into the server and right now you cannot pass CLI arguments to the server to specify the port and host, as the PyTorch stuff also reads from the command line (will fix this soon). If you would like to change this, locate line 195 in backend/server.py.

TODOS

• Change how we run the model, make it easier to use (don't use their options object)
• Make a seperate frontend server and a backend server (for scaling)
• Try to containerize at least the backend components