Basic layout

app.py

@@ -5,55 +5,29 @@ from inference import DepthPredictor, SegmentPredictor
 from utils import create_3d_obj, create_3d_pc, point_cloud
 import numpy as np
 
-def produce_depth_map(image):
-    depth_predictor = DepthPredictor()
-    depth_result = depth_predictor.predict(image)
-    return depth_result
-
-def produce_segmentation_map(image):
-    segment_predictor = SegmentPredictor()
-    sam_result = segment_predictor.predict(image)
-    return sam_result
 
-def produce_3d_reconstruction(image):
+def snap(image, video):
     depth_predictor = DepthPredictor()
     depth_result = depth_predictor.predict(image)
     rgb_gltf_path = create_3d_obj(np.array(image), depth_result, path='./rgb.gltf')
-    return rgb_gltf_path
-
-def produce_point_cloud(depth_map, segmentation_map):
-    return point_cloud(np.array(segmentation_map), depth_map)
-
-def snap(image, depth_map, segmentation_map, video):
-    depth_result = produce_depth_map(image) if depth_map else None
-    sam_result = produce_segmentation_map(image) if segmentation_map else None
-    rgb_gltf_path = produce_3d_reconstruction(image) if depth_map else None
-    point_cloud_fig = produce_point_cloud(depth_result, sam_result) if segmentation_map else None
     
-    if video:
-        # Add video processing here if needed
-        pass
-
-    return [image, depth_result, sam_result, rgb_gltf_path, point_cloud_fig]
+    segment_predictor = SegmentPredictor()
+    sam_result = segment_predictor.predict(image)
+    fig = point_cloud(np.array(sam_result), depth_result)
 
-# Interface inputs
-image_input = gr.Image(source="webcam", tool=None, label="Input Image", type="pil")
-depth_map_button = gr.Button(label="Produce Depth Map", value=False)
-segmentation_map_button = gr.Button(label="Produce Segmentation Map", value=False)
-video_input = gr.Video(source="webcam")
+    return [image, depth_result, sam_result, rgb_gltf_path, fig]#[depth_result, gltf_path, gltf_path]
 
-# Interface outputs
-output_image = gr.Image(label="RGB")
-output_depth_map = gr.Image(label="Predicted Depth")
-output_segmentation_map = gr.Image(label="Predicted Segmentation")
-output_3d_reconstruction = gr.Model3D(label="3D mesh reconstruction - RGB", clear_color=[1.0, 1.0, 1.0, 1.0])
-output_point_cloud = gr.Plot(label="Point Cloud")
 
-# Interface
 demo = gr.Interface(
     snap,
-    inputs=[image_input, depth_map_button, segmentation_map_button, video_input],
-    outputs=[output_image, output_depth_map, output_segmentation_map, output_3d_reconstruction, output_point_cloud]
+    inputs=[gr.Image(source="webcam", tool=None, label="Input Image", type="pil"),
+            gr.Video(source="webcam")],
+    outputs=[gr.Image(label="RGB"),
+             gr.Image(label="predicted depth"),
+             gr.Image(label="predicted segmentation"),
+             gr.Model3D(label="3D mesh reconstruction - RGB",
+                        clear_color=[1.0, 1.0, 1.0, 1.0]),
+             gr.Plot()]
 )
 
 if __name__ == "__main__":