The base ONNX models can be found here.
The scripts for fusing pre/post-processing operations can be found on the conversion_scripts branch.

The models suffixed with _256 take a 256 * 256 sized RGB image as input, whereas other models accept a 512 * 512 sized RGB image.

Read the Depth-Anything-V2 paper

The following release contains model.onnx and model_fp16.onnx models which are to be placed in the app/src/main/assets directory to run the Android app. See README.md for more details.