fix(ml): handle empty/corrupt images in face detection (#27391)

* fix(ml): handle empty/corrupt images in face detection

When a corrupt or degenerate image with zero-dimension (0 width or 0 height)
reaches the face detection pipeline, insightface's RetinaFace.detect() calls
cv2.resize() with a target size of 0, triggering an OpenCV assertion failure:

  error: (-215:Assertion failed) inv_scale_x > 0 in function 'resize'

This crashes the ML worker and returns a 500 error to the server.

Add an early return in FaceDetector._predict() that checks for zero-dimension
images after decoding and returns empty detection results instead of passing
them to the insightface model.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(ml): move empty image validation to request level

Per review feedback, validate image dimensions in the predict endpoint
(returning 400) rather than in each model's _predict method. This
catches all zero-dimension images before they reach any model task.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(ml): resolve mypy strict type error in predict endpoint

Use intermediate `decoded` variable so mypy knows `.width` and `.height`
are accessed on `Image`, not on `Image | str`.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

machine-learning/immich_ml/main.py

@@ -183,7 +183,10 @@ async def predict(
     text: str | None = Form(default=None),
 ) -> Any:
     if image is not None:
-        inputs: Image | str = await run(lambda: decode_pil(image))
+        decoded = await run(lambda: decode_pil(image))
+        if decoded.width == 0 or decoded.height == 0:
+            raise HTTPException(400, "Image has zero width or height")
+        inputs: Image | str = decoded
     elif text is not None:
         inputs = text
     else:

machine-learning/test_main.py

@@ -1198,6 +1198,19 @@ class TestLoad:
         mock_model.model_format = ModelFormat.ONNX
 
 
+@pytest.mark.parametrize("size", [(0, 100), (100, 0), (0, 0)])
+def test_predict_rejects_empty_image(size: tuple[int, int], deployed_app: TestClient) -> None:
+    with mock.patch("immich_ml.main.decode_pil", return_value=Image.new("RGB", size)):
+        response = deployed_app.post(
+            "http://localhost:3003/predict",
+            data={"entries": json.dumps({"clip": {"visual": {"modelName": "ViT-B-32__openai"}}})},
+            files={"image": b"fake image bytes"},
+        )
+
+    assert response.status_code == 400
+    assert "zero" in response.json()["detail"].lower()
+
+
 def test_root_endpoint(deployed_app: TestClient) -> None:
     response = deployed_app.get("http://localhost:3003")