https://ieeexplore.ieee.org/document/9288276
Image generation from sketch is a popular and well-studied computer vision problem. However, the inverse problem image-to-sketch (I2S) synthesis still remains open and challenging, let alone image-to-scene sketch (I2S 2 ) synthesis, especially when full-scene sketch generations are highly desired. This research investigates a framework for generating full-scene sketch representations from natural scene images, aiming to generate outputs that approximate hand-drawn scene sketches.
The base architecture for this project is a pytorch implementation of CycleGAN.
Exploratory work includes:
- Designed Canny edge detection pytorch module for additional input
- Implemented Holistically-Nested Edge Detection architecture for additional input
- Implemented Wasserstein loss metric as alternative to least squares and cross entropy
- Experimented with learning rate decay and other scheduling
Canny Edge Detection:
CycleGAN: Project | Paper | Torch
Pix2pix: Project | Paper | Torch
- PyTorch 0.41+
- opencv-python
See environment.yml
- Dan McGonigle dpmcgonigle