An application to automatically remove selected objects from images. It combines image segmentation with image inpainting to create automated pipeline to remove whole objects. There are two removal modes:
- all instance removal (semantic segmentation is used)
- single instance removal (instance segmentation is used)
Implemented segmentation models:
- Deeplab (torchvision weigths used)
Used image inpainting models:
- CR fill (this repository)
The project is WIP.
- Setup and enter pipenv environment
pipenv sync
pipenv shell- Run the program
- Semantic segmentation (all instances of selected classes are removed)
python app.py --mode semseg_imgExamples:
| Image | Video |
|---|---|
- Instance segmentation (only selected instances are removed)
python app.py --mode instseg_imgExamples:
| Image | Video |
|---|---|