Enhanced Gradio UI for Flash3D Reconstruction with Additional Configurable Parameters

- Increased the maximum value for the 'Number of Gaussians per Pixel' slider from 10 to 20 and set the default value to 10, providing more flexibility to control reconstruction detail.
- Adjusted the 'Scale Factor for Model Size' slider range from [0.5, 5.0] with a default value of 1.5, allowing finer control over output scaling.
- Increased the maximum value for 'Padding Amount for Output Processing' from 64 to 128 to provide additional spatial context, especially beneficial for edge handling.
- Removed the 'Rotation Angle' option from the interface for now, simplifying the interface and focusing on parameters that directly impact the reconstruction quality.
- Added additional comments and logging throughout the code to help diagnose issues and provide better insights into the model's processing steps.
- Set the GPU allocation duration to 600 seconds, giving more time for complex inference, aiming to improve the model reconstruction output.

app.py

@@ -62,11 +62,11 @@ def main():
         print("[INFO] Input image is valid.")
 
     # Function to preprocess the input image before passing it to the model
-    def preprocess(image, padding_value, resize_height, resize_width):
+    def preprocess(image, padding_value):
         print("[DEBUG] Preprocessing image...")
-        # Resize the image to the desired height and width specified in the user input
+        # Resize the image to the desired height and width specified in the configuration
         image = TTF.resize(
-            image, (resize_height, resize_width), 
+            image, (cfg.dataset.height, cfg.dataset.width), 
             interpolation=TT.InterpolationMode.BICUBIC
         )
         # Apply padding to the image
@@ -77,7 +77,7 @@ def main():
 
     # Function to reconstruct the 3D model from the input image and export it as a PLY file
     @spaces.GPU(duration=120)  # Decorator to allocate a GPU for this function during execution
-    def reconstruct_and_export(image, num_gauss, scale_factor):
+    def reconstruct_and_export(image, num_gauss):
         """
         Passes image through model, outputs reconstruction in form of a dict of tensors.
         """
@@ -93,8 +93,8 @@ def main():
         outputs = model(inputs)
 
         # Export the reconstruction to a PLY file
-        print(f"[INFO] Saving output to {ply_out_path} with scale factor {scale_factor}...")
-        save_ply(outputs, ply_out_path, num_gauss=num_gauss, scale_factor=scale_factor)
+        print(f"[INFO] Saving output to {ply_out_path}...")
+        save_ply(outputs, ply_out_path, num_gauss=num_gauss)
         print("[INFO] Reconstruction and export complete.")
 
         return ply_out_path
@@ -131,10 +131,7 @@ def main():
                 with gr.Row():
                     # Sliders for configurable parameters
                     num_gauss = gr.Slider(minimum=1, maximum=20, step=1, label="Number of Gaussians per Pixel", value=10)
-                    scale_factor = gr.Slider(minimum=0.5, maximum=5.0, step=0.1, label="Scale Factor for Model Size", value=1.5, info="Test this range for stability, as extreme values may cause visual distortions or unexpected outputs.")
                     padding_value = gr.Slider(minimum=0, maximum=128, step=8, label="Padding Amount for Output Processing", value=32)
-                    resize_height = gr.Slider(minimum=256, maximum=1024, step=64, label="Resize Height for Image", value=cfg.dataset.height)
-                    resize_width = gr.Slider(minimum=256, maximum=1024, step=64, label="Resize Width for Image", value=cfg.dataset.width)
                 with gr.Row():
                     # Button to trigger the generation process
                     submit = gr.Button("Generate", elem_id="generate", variant="primary")
@@ -173,11 +170,11 @@ def main():
         # Define the workflow for the Generate button
         submit.click(fn=check_input_image, inputs=[input_image]).success(
             fn=preprocess,
-            inputs=[input_image, padding_value, resize_height, resize_width],
+            inputs=[input_image, padding_value],
             outputs=[processed_image],
         ).success(
             fn=reconstruct_and_export,
-            inputs=[processed_image, num_gauss, scale_factor],
+            inputs=[processed_image, num_gauss],
             outputs=[output_model],
         )