attempt to fix hf space setup

Dockerfile

@@ -4,7 +4,7 @@ FROM nvidia/cuda:11.7.1-devel-ubuntu22.04
 ENV LC_ALL C.UTF-8
 ENV LANG C.UTF-8
 ARG UID=1000
-ARG UNAME=ubuntu
+ARG UNAME=user
 ARG DEBIAN_FRONTEND=noninteractive
 ENV PATH="/home/$UNAME/.local/bin:$PATH"
 
@@ -58,4 +58,4 @@ ENV PYTHONPATH="/usr/local/lib/python3.10/dist-packages/MultiScaleDeformableAtte
 
 WORKDIR /home/$UNAME/opdmulti-demo
 
-RUN python3.10 app.py
+CMD ["python3.10", "app.py"]

README.md

@@ -4,7 +4,6 @@ emoji: 🌍
 colorFrom: gray
 colorTo: red
 sdk: docker
-app_port: 7860
 pinned: false
 license: mit
 ---

app.py

@@ -174,103 +174,106 @@ def clear_outputs():
     return [gr.update(value=None, visible=(idx == 0)) for idx in range(MAX_PARTS)]
 
 
-print("Starting up app...")
-
-with gr.Blocks() as demo:
-    gr.Markdown(
+def run():
+    with gr.Blocks() as demo:
+        gr.Markdown(
+            """
+        # OPDMulti Demo
+        We tackle the openable-part-detection (OPD) problem where we identify in a single-view image parts that are openable and their motion parameters. Our OPDFORMER architecture outputs segmentations for openable parts on potentially multiple objects, along with each part’s motion parameters: motion type (translation or rotation, indicated by blue or purple mask), motion axis and origin (see green arrows and points). For each openable part, we predict the motion parameters (axis and origin) in object coordinates along with an object pose prediction to convert to camera coordinates.
+    
+        More information about the project, including code, can be found [here](https://3dlg-hcvc.github.io/OPDMulti/).
+    
+        Upload an image to see a visualization of its range of motion below. Only the RGB image is needed for the model itself, but the depth image is required as of now for the visualization of motion.
+    
+        If you know the intrinsic matrix of your camera, you can specify that here or otherwise use the default matrix which will work with any of the provided examples.
+    
+        You can also change the number of samples to define the number of states in the visualization generated.
         """
-    # OPDMulti Demo
-    We tackle the openable-part-detection (OPD) problem where we identify in a single-view image parts that are openable and their motion parameters. Our OPDFORMER architecture outputs segmentations for openable parts on potentially multiple objects, along with each part’s motion parameters: motion type (translation or rotation, indicated by blue or purple mask), motion axis and origin (see green arrows and points). For each openable part, we predict the motion parameters (axis and origin) in object coordinates along with an object pose prediction to convert to camera coordinates.
-
-    More information about the project, including code, can be found [here](https://3dlg-hcvc.github.io/OPDMulti/).
-
-    Upload an image to see a visualization of its range of motion below. Only the RGB image is needed for the model itself, but the depth image is required as of now for the visualization of motion.
-
-    If you know the intrinsic matrix of your camera, you can specify that here or otherwise use the default matrix which will work with any of the provided examples.
-
-    You can also change the number of samples to define the number of states in the visualization generated.
-    """
-    )
-
-    # inputs
-    with gr.Row():
-        rgb_image = gr.Image(
-            image_mode="RGB", source="upload", type="filepath", label="RGB Image", show_label=True, interactive=True
-        )
-        depth_image = gr.Image(
-            image_mode="I;16", source="upload", type="filepath", label="Depth Image", show_label=True, interactive=True
         )
-
-    intrinsic = gr.Dataframe(
-        value=[
-            [
-                214.85935872395834,
-                0.0,
-                125.90160319010417,
-            ],
-            [
-                0.0,
-                214.85935872395834,
-                95.13726399739583,
+    
+        # inputs
+        with gr.Row():
+            rgb_image = gr.Image(
+                image_mode="RGB", source="upload", type="filepath", label="RGB Image", show_label=True, interactive=True
+            )
+            depth_image = gr.Image(
+                image_mode="I;16", source="upload", type="filepath", label="Depth Image", show_label=True, interactive=True
+            )
+    
+        intrinsic = gr.Dataframe(
+            value=[
+                [
+                    214.85935872395834,
+                    0.0,
+                    125.90160319010417,
+                ],
+                [
+                    0.0,
+                    214.85935872395834,
+                    95.13726399739583,
+                ],
+                [
+                    0.0,
+                    0.0,
+                    1.0,
+                ],
             ],
-            [
-                0.0,
-                0.0,
-                1.0,
+            row_count=(3, "fixed"),
+            col_count=(3, "fixed"),
+            datatype="number",
+            type="numpy",
+            label="Intrinsic matrix",
+            show_label=True,
+            interactive=True,
+        )
+        num_samples = gr.Number(
+            value=NUM_SAMPLES,
+            label="Number of samples",
+            show_label=True,
+            interactive=True,
+            precision=0,
+            minimum=3,
+            maximum=20,
+        )
+    
+        # specify examples which can be used to start
+        examples = gr.Examples(
+            examples=[
+                ["examples/59-4860.png", "examples/59-4860_d.png"],
+                ["examples/174-8460.png", "examples/174-8460_d.png"],
+                ["examples/187-0.png", "examples/187-0_d.png"],
+                ["examples/187-23040.png", "examples/187-23040_d.png"],
             ],
-        ],
-        row_count=(3, "fixed"),
-        col_count=(3, "fixed"),
-        datatype="number",
-        type="numpy",
-        label="Intrinsic matrix",
-        show_label=True,
-        interactive=True,
-    )
-    num_samples = gr.Number(
-        value=NUM_SAMPLES,
-        label="Number of samples",
-        show_label=True,
-        interactive=True,
-        precision=0,
-        minimum=3,
-        maximum=20,
-    )
-
-    # specify examples which can be used to start
-    examples = gr.Examples(
-        examples=[
-            ["examples/59-4860.png", "examples/59-4860_d.png"],
-            ["examples/174-8460.png", "examples/174-8460_d.png"],
-            ["examples/187-0.png", "examples/187-0_d.png"],
-            ["examples/187-23040.png", "examples/187-23040_d.png"],
-        ],
-        inputs=[rgb_image, depth_image],
-        api_name=False,
-        examples_per_page=2,
-    )
-
-    submit_btn = gr.Button("Run model")
-
-    # output
-    explanation = gr.Markdown(
-        value=f"# Output\nClick on an image to see an animation of the part motion. As of now, only up to {MAX_PARTS} parts can be visualized due to limitations of the visualizer."
-    )
-
-    images = [
-        gr.Image(type="pil", label=f"Part {idx + 1}", show_download_button=False, visible=(idx == 0))
-        for idx in range(MAX_PARTS)
-    ]
-    for idx, image_comp in enumerate(images):
-        image_comp.select(get_trigger(idx), inputs=rgb_image, outputs=image_comp, api_name=False)
-
-    # if user changes input, clear output images
-    rgb_image.change(clear_outputs, inputs=[], outputs=images, api_name=False)
-    depth_image.change(clear_outputs, inputs=[], outputs=images, api_name=False)
-
-    submit_btn.click(
-        fn=predict, inputs=[rgb_image, depth_image, intrinsic, num_samples], outputs=images, api_name=False
-    )
+            inputs=[rgb_image, depth_image],
+            api_name=False,
+            examples_per_page=2,
+        )
+    
+        submit_btn = gr.Button("Run model")
+    
+        # output
+        explanation = gr.Markdown(
+            value=f"# Output\nClick on an image to see an animation of the part motion. As of now, only up to {MAX_PARTS} parts can be visualized due to limitations of the visualizer."
+        )
+    
+        images = [
+            gr.Image(type="pil", label=f"Part {idx + 1}", show_download_button=False, visible=(idx == 0))
+            for idx in range(MAX_PARTS)
+        ]
+        for idx, image_comp in enumerate(images):
+            image_comp.select(get_trigger(idx), inputs=rgb_image, outputs=image_comp, api_name=False)
+    
+        # if user changes input, clear output images
+        rgb_image.change(clear_outputs, inputs=[], outputs=images, api_name=False)
+        depth_image.change(clear_outputs, inputs=[], outputs=images, api_name=False)
+    
+        submit_btn.click(
+            fn=predict, inputs=[rgb_image, depth_image, intrinsic, num_samples], outputs=images, api_name=False
+        )
+    
+    demo.queue(api_open=False)
+    demo.launch(server_name="0.0.0.0", server_port=7860)
 
-demo.queue(api_open=False)
-demo.launch()
+if __name__ == "__main__":
+    print("Starting up app...")
+    run()