| --- |
| license: apache-2.0 |
| pipeline_tag: depth-estimation |
| library_name: executorch |
| base_model: depth-anything/Depth-Anything-V2-Small-hf |
| --- |
| |
| # depth-anything-v2 |
|
|
| This repository hosts **Depth Anything V2 Small** (ViT-S/14) exported for |
| [React Native ExecuTorch](https://www.npmjs.com/package/react-native-executorch) |
| as ExecuTorch `.pte` programs for on-device relative depth estimation. |
|
|
| Upstream model: [depth-anything/Depth-Anything-V2-Small-hf](https://huggingface.co/depth-anything/Depth-Anything-V2-Small-hf), |
| revision `5426e4f0f36572d16453bbda7a8389317b1bef99`. |
|
|
| ## Variants |
|
|
| | Path | Backend | Precision | |
| | --- | --- | --- | |
| | `small/xnnpack/depth_anything_v2_small_xnnpack_fp32.pte` | xnnpack | fp32 | |
| | `small/xnnpack/depth_anything_v2_small_xnnpack_int8.pte` | xnnpack | int8 | |
| | `small/coreml/depth_anything_v2_small_coreml_fp16.pte` | coreml | fp16 | |
| | `small/vulkan/depth_anything_v2_small_vulkan_fp16.pte` | vulkan | fp16 | |
|
|
| ## XNNPACK int8 |
|
|
| The INT8 export uses dynamic INT8 activations and per-channel INT8 weights for |
| linear layers; convolutions remain floating point. Public inputs and outputs |
| remain float32. |
|
|
| Numerical agreement with the float reference on four held-out upstream example |
| illustrations (`demo17`–`demo20`): |
|
|
| | Image | Range-normalized RMSE | Correlation | |
| | --- | --- | --- | |
| | demo17 | 1.24% | 0.999633 | |
| | demo18 | 1.34% | 0.999104 | |
| | demo19 | 0.47% | 0.999970 | |
| | demo20 | 1.28% | 0.999690 | |
|
|
| Range-normalized RMSE is RMSE divided by the reference depth range. These are |
| numerical export checks, not depth-accuracy benchmarks. A synthetic random-input |
| check reached 3.32% range-normalized RMSE (correlation 0.990703). |
|
|
| ## Repository structure |
|
|
| ``` |
| config.json 39 B |
| small/coreml/config.json 879 B |
| small/coreml/depth_anything_v2_small_coreml_fp16.pte 50.2 MB |
| small/vulkan/config.json 879 B |
| small/vulkan/depth_anything_v2_small_vulkan_fp16.pte 49.6 MB |
| small/xnnpack/config.json 1.5 kB |
| small/xnnpack/depth_anything_v2_small_xnnpack_fp32.pte 99.0 MB |
| small/xnnpack/depth_anything_v2_small_xnnpack_int8.pte 35.5 MB |
| ``` |
|
|
| ## Compatibility |
|
|
| These files are published for the **ExecuTorch v1.4.1** runtime. ExecuTorch gives |
| no forward compatibility guarantee, so an older runtime may fail to load them. |
|
|
| For React Native ExecuTorch model loading, see the |
| [documentation](https://docs.swmansion.com/react-native-executorch/docs/fundamentals/downloading-models). |
| To load these files in your own runtime, read the |
| [compatibility note](https://github.com/pytorch/executorch/blob/main/runtime/COMPATIBILITY.md). |
|
|
