knazeri/edge-connect
A two-stage adversarial model that generates missing image regions by first hallucinating edges and then completing the image.
Not currently ranked — collecting fresh signals.
star history
EdgeConnect is an image inpainting method published at ICCV 2019 that follows a lines-first-then-color approach. The model uses a two-stage architecture: an edge generator hallucinates missing edges for damaged regions, and a subsequent image completion network fills in the missing pixels using these hallucinated edges as guidance. Both stages use adversarial training with GANs and are implemented in PyTorch.
Frequently asked
- What is knazeri/edge-connect?
- A two-stage adversarial model that generates missing image regions by first hallucinating edges and then completing the image.
- Is edge-connect open source?
- Yes — knazeri/edge-connect is an open-source project tracked on heatdrop.
- What language is edge-connect written in?
- knazeri/edge-connect is primarily written in Python.
- How popular is edge-connect?
- knazeri/edge-connect has 2.6k stars on GitHub.
- Where can I find edge-connect?
- knazeri/edge-connect is on GitHub at https://github.com/knazeri/edge-connect.