Version 0.2.0

Acly released this 17 Oct 12:00

· 18 commits to main since this release

25dc418

Depth Anything V2 inference
Public API to reuse backbones: SWINv1, DINOv2
BiRefNet: enabled Flash Attention (improves performance on GPU)
BiRefNet: automatically resize dynamic inputs to fit maximum GPU buffer size
Extended visp::backend_type enum to cover specific GPU backend in the future
Fixed support for integrated GPUs

Assets 4