LaMa:
Resolution-robust Large Mask Inpainting with Fourier Convolutions, WACV 2022 by Samsung Research. Alternatively, for automatic image inpainting LaMa with MaskDINO.
DeOldify:
A Deep Learning based project for colorizing and restoring old images.
DIS (v2 upcoming):
Highly Accurate Dichotomous Image Segmentation from the maker of U2-Net.
Those are very nice models and definitely deserve to be added, but I can only work with low-memory models that my potato-specs notebook can handle. Also DIS is kinda difficult to add due to how it's made...
Please add these powerful general uses AI models