Aesthetic Shadow V2 is a 1.1b parameters visual transformer designed to evaluate the quality of anime images. It accepts high-resolution 1024x1024 images as input and provides a prediction score that quantifies the aesthetic appeal of the artwork. Leveraging cutting-edge deep learning techniques, this model excels at discerning fine details, proportions, and overall visual coherence in anime illustrations.

This is an improved model over the original shadowlilac/aesthetic-shadow

Disclosure

This model does not intend to be offensive towards any artist and may not output an accurate label for an image. A potential use case would be low quality images filtering on image datasets.

Downloads last month
37
Safetensors
Model size
1.09B params
Tensor type
F32
·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.