Acly/Depth-Anything-V2-GGUF

GGUF models for Depth-Anything V2

Depth-Anything is a model for monocular depth estimation. The weights in this repository are converted for lightweight inference on consumer hardware with vision.cpp.

• Original repository: DepthAnything/Depth-Anything-V2 (Github)
• Original weights: depth-anything (HuggingFace)

Run

Example inference with vision.cpp:

CLI

vision-cli depth-anything -m Depth-Anything-V2-Small-F16.gguf -i input.png -o depth.png

C++

image_data     image  = image_load("input.png");
backend_device device = backend_init();
depthany_model model  = depthany_load_model("Depth-Anything-V2-Small-F16.gguf", device);
image_data     depth  = depthany_compute(model, image);
image_data     output = image_f32_to_u8(depth, image_format::alpha_u8);
image_save(output, "depth.png");

License

• Depth-Anything-V2-Small: Apache-2.0
• Depth-Anything-V2-Base: CC-BY-NC-4.0
• Depth-Anything-V2-Large: CC-BY-NC-4.0