Autoencoder in CIELAB color space
- LAB input, RGB output
- Scaling, Rotational equivalency
- B, C, H, W shape
After many retries at creating small decoders for VAE, rotational transformations in VQ-VAE, in HSL, HSV, YCbCr and RGB color spaces; here's the LAB variant.
The latent representation should not produce artifacts under rotations or hue changes. However, it does not obey reflection symmetry.
This model was made for further analysis of traditional art pieces.
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
🙋
Ask for provider support