fix(client): clip float images before convert_to_uint8

packages/openpi-client/src/openpi_client/image_tools.py

@@ -6,9 +6,12 @@ def convert_to_uint8(img: np.ndarray) -> np.ndarray:
     """Converts an image to uint8 if it is a float image.
 
     This is important for reducing the size of the image when sending it over the network.
+
+    Float inputs are assumed to be in approximately the [0, 1] range. Values are
+    clipped before scaling so out-of-range pixels do not wrap under `astype(uint8)`.
     """
     if np.issubdtype(img.dtype, np.floating):
-        img = (255 * img).astype(np.uint8)
+        img = (255 * np.clip(img, 0.0, 1.0)).astype(np.uint8)
     return img
 
 

packages/openpi-client/src/openpi_client/image_tools_test.py

@@ -35,3 +35,17 @@ def test_resize_with_pad_shapes():
     resized_images = image_tools.resize_with_pad(images, height, width)
     assert resized_images.shape == (1, height, width, 3)
     assert np.all(resized_images == 0)
+
+
+def test_convert_to_uint8_clips_out_of_range_floats():
+    img = np.array([[[-0.5, 0.0, 1.5]]], dtype=np.float32)
+    out = image_tools.convert_to_uint8(img)
+    assert out.dtype == np.uint8
+    assert out.tolist() == [[[0, 0, 255]]]
+
+
+def test_convert_to_uint8_passes_through_uint8():
+    img = np.array([[[10, 20, 30]]], dtype=np.uint8)
+    out = image_tools.convert_to_uint8(img)
+    assert out.dtype == np.uint8
+    assert np.array_equal(out, img)