Add Depth Anything V2 Small ExecuTorch models

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.pte filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,3 +1,77 @@
 ---
 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 |
+
+## Input and output
+
+The `forward` method takes a float32 RGB tensor `[1, 3, 518, 518]` in NCHW
+layout. Resize the image to 518 × 518 using bicubic interpolation, scale pixel
+values to `[0, 1]`, then normalize channels with mean `[0.485, 0.456, 0.406]`
+and standard deviation `[0.229, 0.224, 0.225]`.
+
+The output is a float32 tensor `[1, 518, 518]` containing raw relative inverse
+depth: larger values indicate closer regions. Values are not distances in
+meters and are not normalized to `[0, 1]`. Resize the depth map to the source
+image dimensions as needed. All variants retain float32 public inputs and outputs.
+
+## XNNPACK int8
+
+The INT8 variant uses per-channel INT8 weights and dynamic INT8 activation
+quantization for linear layers; convolutions remain floating point. It is a
+mixed-precision model. FP32 remains the default variant in the XNNPACK config.
+
+INT8 passed a provisional check of range-normalized RMSE below 2% and
+correlation above 0.99 on four held-out upstream example images. A synthetic
+random input exceeded that RMSE threshold (3.32%). This is limited numerical
+validation, not a depth benchmark or a task accuracy guarantee.
+
+## Backend validation
+
+- XNNPACK FP32 and INT8 were executed on macOS and checked against the float model.
+- CoreML FP16 was executed on macOS and numerically checked against the float model.
+- Vulkan FP16 was exported and its serialized contract and delegation inspected;
+  execution on a Vulkan device has not been validated.
+
+## Repository structure
+
+```
+config.json                                                   39 B
+small/coreml/config.json                                      928 B
+small/coreml/depth_anything_v2_small_coreml_fp16.pte          50.2 MB
+small/vulkan/config.json                                      928 B
+small/vulkan/depth_anything_v2_small_vulkan_fp16.pte          49.6 MB
+small/xnnpack/config.json                                     1.6 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).

config.json

@@ -0,0 +1,3 @@
+{
+  "modelName": "depth-anything-v2"
+}

small/coreml/config.json

@@ -0,0 +1,43 @@
+{
+  "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
+  "model": "depth_anything_v2_small",
+  "family": "depth-anything-v2",
+  "capabilities": [
+    "depth-estimation"
+  ],
+  "backend": "coreml",
+  "license": "apache-2.0",
+  "variants": [
+    {
+      "file": "depth_anything_v2_small_coreml_fp16.pte",
+      "precision": "fp16",
+      "quantized": false,
+      "default": true,
+      "methods": {
+        "forward": {
+          "inputs": [
+            {
+              "shape": [
+                1,
+                3,
+                518,
+                518
+              ],
+              "dtype": "float32"
+            }
+          ],
+          "outputs": [
+            {
+              "shape": [
+                1,
+                518,
+                518
+              ],
+              "dtype": "float32"
+            }
+          ]
+        }
+      }
+    }
+  ]
+}

small/coreml/depth_anything_v2_small_coreml_fp16.pte

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e6e06eee9067b2b884b8f79f9444313db596704e2baf9f02d2eda3dde628071
+size 50181365

small/vulkan/config.json

@@ -0,0 +1,43 @@
+{
+  "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
+  "model": "depth_anything_v2_small",
+  "family": "depth-anything-v2",
+  "capabilities": [
+    "depth-estimation"
+  ],
+  "backend": "vulkan",
+  "license": "apache-2.0",
+  "variants": [
+    {
+      "file": "depth_anything_v2_small_vulkan_fp16.pte",
+      "precision": "fp16",
+      "quantized": false,
+      "default": true,
+      "methods": {
+        "forward": {
+          "inputs": [
+            {
+              "shape": [
+                1,
+                3,
+                518,
+                518
+              ],
+              "dtype": "float32"
+            }
+          ],
+          "outputs": [
+            {
+              "shape": [
+                1,
+                518,
+                518
+              ],
+              "dtype": "float32"
+            }
+          ]
+        }
+      }
+    }
+  ]
+}

small/vulkan/depth_anything_v2_small_vulkan_fp16.pte

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aed86d706e97a37db1bb54b385e350dde806ce0507b9465779b4bcd532998790
+size 49595524

small/xnnpack/config.json

@@ -0,0 +1,74 @@
+{
+  "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
+  "model": "depth_anything_v2_small",
+  "family": "depth-anything-v2",
+  "capabilities": [
+    "depth-estimation"
+  ],
+  "backend": "xnnpack",
+  "license": "apache-2.0",
+  "variants": [
+    {
+      "file": "depth_anything_v2_small_xnnpack_fp32.pte",
+      "precision": "fp32",
+      "quantized": false,
+      "default": true,
+      "methods": {
+        "forward": {
+          "inputs": [
+            {
+              "shape": [
+                1,
+                3,
+                518,
+                518
+              ],
+              "dtype": "float32"
+            }
+          ],
+          "outputs": [
+            {
+              "shape": [
+                1,
+                518,
+                518
+              ],
+              "dtype": "float32"
+            }
+          ]
+        }
+      }
+    },
+    {
+      "file": "depth_anything_v2_small_xnnpack_int8.pte",
+      "precision": "int8",
+      "quantized": true,
+      "default": false,
+      "methods": {
+        "forward": {
+          "inputs": [
+            {
+              "shape": [
+                1,
+                3,
+                518,
+                518
+              ],
+              "dtype": "float32"
+            }
+          ],
+          "outputs": [
+            {
+              "shape": [
+                1,
+                518,
+                518
+              ],
+              "dtype": "float32"
+            }
+          ]
+        }
+      }
+    }
+  ]
+}

small/xnnpack/depth_anything_v2_small_xnnpack_fp32.pte

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f83737200f4421a898c5b975695dacf4c693b0546d718adf96e9e6645084718f
+size 99023364

small/xnnpack/depth_anything_v2_small_xnnpack_int8.pte

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3ccc836262d16529fd91b734c5df35da55169a905fcf7958992fc71b0c74ecb
+size 35481732