Update app.py

app.py

@@ -27,6 +27,8 @@ from src.model.model.anysplat import AnySplat
 from src.model.ply_export import export_ply
 from src.utils.image import process_image
 
+os.environ["ANYSPLAT_PROCESSED"] = f"{os.getcwd()}/proprocess_results"
+
 
 # 1) Core model inference
 def get_reconstructed_scene(outdir, model, device):
@@ -77,7 +79,7 @@ def get_reconstructed_scene(outdir, model, device):
 
 
 # 2) Handle uploaded video/images --> produce target_dir + images
-def handle_uploads(input_video, input_images):
+def handle_uploads(input_video, input_images, session_id):
     """
     Create a new 'target_dir' + 'images' subfolder, and place user-uploaded
     images or extracted frames from video into it. Return (target_dir, image_paths).
@@ -86,9 +88,8 @@ def handle_uploads(input_video, input_images):
     gc.collect()
     torch.cuda.empty_cache()
 
-    # Create a unique folder name
-    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S_%f")
-    target_dir = f"input_images_{timestamp}"
+    base_dir = os.path.join(os.environ["ANYSPLAT_PROCESSED"], session_id)
+    target_dir = base_dir
     target_dir_images = os.path.join(target_dir, "images")
 
     # Clean up if somehow that folder already exists
@@ -160,34 +161,24 @@ def update_gallery_on_upload(input_video, input_images):
 
 
 @spaces.GPU()
-# 4) Reconstruction: uses the target_dir plus any viz parameters
-def gradio_demo(
-    target_dir,
-):
-    """
-    Perform reconstruction using the already-created target_dir/images.
-    """
-    if not os.path.isdir(target_dir) or target_dir == "None":
-        return None, None, None
+def generate_splat(images_folder, session_id=None):
 
     start_time = time.time()
     gc.collect()
     torch.cuda.empty_cache()
-    
-    # Prepare frame_filter dropdown
-    target_dir_images = os.path.join(target_dir, "images")
+
+    base_dir = os.path.join(os.environ["ANYSPLAT_PROCESSED"], session_id)
+
     all_files = (
-        sorted(os.listdir(target_dir_images))
-        if os.path.isdir(target_dir_images)
+        sorted(os.listdir(images_folder))
+        if os.path.isdir(images_folder)
         else []
     )
     all_files = [f"{i}: {filename}" for i, filename in enumerate(all_files)]
 
     print("Running run_model...")
     with torch.no_grad():
-        plyfile, video, depth_colored = get_reconstructed_scene(
-            target_dir, model, device
-        )
+        plyfile, video, depth_colored = get_reconstructed_scene(base_dir, model, device)
 
     end_time = time.time()
     print(f"Total time: {end_time - start_time:.2f} seconds (including IO)")
@@ -195,11 +186,21 @@ def gradio_demo(
     return plyfile, video, depth_colored
 
 
-def clear_fields():
+def start_session(request: gr.Request):
     """
-    Clears the 3D viewer, the stored target_dir, and empties the gallery.
+    Initialize a new user session and return the session identifier.
+    
+    This function is triggered when the Gradio demo loads and creates a unique
+    session hash that will be used to organize outputs and temporary files
+    for this specific user session.
+    
+    Args:
+        request (gr.Request): Gradio request object containing session information
+        
+    Returns:
+        str: Unique session hash identifier
     """
-    return None, None, None
+    return request.session_hash
 
 
 if __name__ == "__main__":
@@ -264,8 +265,9 @@ if __name__ == "__main__":
         }
         """
     with gr.Blocks(css=css, title="AnySplat Demo", theme=theme) as demo:
-
-            
+        session_state = gr.State()
+        demo.load(start_session, outputs=[session_state])
+                
         target_dir_output = gr.Textbox(label="Target Dir", visible=False, value="None")
         is_example = gr.Textbox(label="is_example", visible=False, value="None")
         num_images = gr.Textbox(label="num_images", visible=False, value="None")
@@ -275,7 +277,6 @@ if __name__ == "__main__":
 
         with gr.Column(elem_id="col-container"):
 
-            
             gr.Markdown(
                 """ # AnySplat – Feed-forward 3D Gaussian Splatting from Unconstrained Views
         
@@ -329,92 +330,60 @@ if __name__ == "__main__":
 
             # ---------------------- Examples section ----------------------
     
-            examples = [
-                [None, "examples/video/re10k_1eca36ec55b88fe4.mp4", "re10k", "1eca36ec55b88fe4", "2", "Real", "True",],
-                [None, "examples/video/bungeenerf_colosseum.mp4", "bungeenerf", "colosseum", "8", "Synthetic", "True",],
-                [None, "examples/video/fox.mp4", "InstantNGP", "fox", "14", "Real", "True",],
-                [None, "examples/video/matrixcity_street.mp4", "matrixcity", "street", "32", "Synthetic", "True",],
-                [None, "examples/video/vrnerf_apartment.mp4", "vrnerf", "apartment", "32", "Real", "True",],
-                [None, "examples/video/vrnerf_kitchen.mp4", "vrnerf", "kitchen", "17", "Real", "True",],
-                [None, "examples/video/vrnerf_riverview.mp4", "vrnerf", "riverview", "12", "Real", "True",],
-                [None, "examples/video/vrnerf_workshop.mp4", "vrnerf", "workshop", "32", "Real", "True",],
-                [None, "examples/video/fillerbuster_ramen.mp4", "fillerbuster", "ramen", "32", "Real", "True",],
-                [None, "examples/video/meganerf_rubble.mp4", "meganerf", "rubble", "10", "Real", "True",],
-                [None, "examples/video/llff_horns.mp4", "llff", "horns", "12", "Real", "True",],
-                [None, "examples/video/llff_fortress.mp4", "llff", "fortress", "7", "Real", "True",],
-                [None, "examples/video/dtu_scan_106.mp4", "dtu", "scan_106", "20", "Real", "True",],
-                [None, "examples/video/horizongs_hillside_summer.mp4", "horizongs", "hillside_summer", "55", "Synthetic", "True",],
-                [None, "examples/video/kitti360.mp4", "kitti360", "kitti360", "64", "Real", "True",],
-            ]
-    
-            def example_pipeline(
-                input_images,
-                input_video,
-                dataset_name,
-                scene_name,
-                num_images_str,
-                image_type,
-                is_example,
-            ):
-                """
-                1) Copy example images to new target_dir
-                2) Reconstruct
-                3) Return model3D + logs + new_dir + updated dropdown + gallery
-                We do NOT return is_example. It's just an input.
-                """
-                target_dir, image_paths = handle_uploads(input_video, input_images)
-                plyfile, video, depth_colored = gradio_demo(target_dir)
-                return plyfile, video, depth_colored, target_dir, image_paths
+            # examples = [
+            #     [None, "examples/video/re10k_1eca36ec55b88fe4.mp4", "re10k", "1eca36ec55b88fe4", "2", "Real", "True",],
+            #     [None, "examples/video/bungeenerf_colosseum.mp4", "bungeenerf", "colosseum", "8", "Synthetic", "True",],
+            #     [None, "examples/video/fox.mp4", "InstantNGP", "fox", "14", "Real", "True",],
+            #     [None, "examples/video/matrixcity_street.mp4", "matrixcity", "street", "32", "Synthetic", "True",],
+            #     [None, "examples/video/vrnerf_apartment.mp4", "vrnerf", "apartment", "32", "Real", "True",],
+            #     [None, "examples/video/vrnerf_kitchen.mp4", "vrnerf", "kitchen", "17", "Real", "True",],
+            #     [None, "examples/video/vrnerf_riverview.mp4", "vrnerf", "riverview", "12", "Real", "True",],
+            #     [None, "examples/video/vrnerf_workshop.mp4", "vrnerf", "workshop", "32", "Real", "True",],
+            #     [None, "examples/video/fillerbuster_ramen.mp4", "fillerbuster", "ramen", "32", "Real", "True",],
+            #     [None, "examples/video/meganerf_rubble.mp4", "meganerf", "rubble", "10", "Real", "True",],
+            #     [None, "examples/video/llff_horns.mp4", "llff", "horns", "12", "Real", "True",],
+            #     [None, "examples/video/llff_fortress.mp4", "llff", "fortress", "7", "Real", "True",],
+            #     [None, "examples/video/dtu_scan_106.mp4", "dtu", "scan_106", "20", "Real", "True",],
+            #     [None, "examples/video/horizongs_hillside_summer.mp4", "horizongs", "hillside_summer", "55", "Synthetic", "True",],
+            #     [None, "examples/video/kitti360.mp4", "kitti360", "kitti360", "64", "Real", "True",],
+            # ]
     
-            gr.Examples(
-                examples=examples,
-                inputs=[
-                    input_images,
-                    input_video,
-                    dataset_name,
-                    scene_name,
-                    num_images,
-                    image_type,
-                    is_example,
-                ],
-                outputs=[
-                    reconstruction_output,
-                    rgb_video,
-                    depth_video,
-                    target_dir_output,
-                    image_gallery,
-                ],
-                fn=example_pipeline,
-                cache_examples=False,
-                examples_per_page=50,
-            )
-
-            gr.Markdown("<p style='text-align: center; font-style: italic; color: #666;'>We thank VGGT for their excellent gradio implementation!</p>")
-
+            # gr.Examples(
+            #     examples=examples,
+            #     inputs=[
+            #         input_images,
+            #         input_video,
+            #         dataset_name,
+            #         scene_name,
+            #         num_images,
+            #         image_type,
+            #         is_example,
+            #     ],
+            #     outputs=[
+            #         reconstruction_output,
+            #         rgb_video,
+            #         depth_video,
+            #         target_dir_output,
+            #         image_gallery,
+            #     ],
+            #     fn=example_pipeline,
+            #     cache_examples=False,
+            #     examples_per_page=50,
+            # )
+            
         submit_btn.click(
-            fn=clear_fields,
-            inputs=[],
-            outputs=[reconstruction_output, rgb_video, depth_video],
-        ).then(
-            fn=gradio_demo,
-            inputs=[
-                target_dir_output,
-            ],
-            outputs=[reconstruction_output, rgb_video, depth_video],
-        ).then(
-            fn=lambda: "False", inputs=[], outputs=[is_example]
-        )
+            fn=generate_splat,
+            inputs=[target_dir_output,],
+            outputs=[reconstruction_output, rgb_video, depth_video])
 
         input_video.change(
             fn=update_gallery_on_upload,
-            inputs=[input_video, input_images],
+            inputs=[input_video, input_images, session_state],
             outputs=[reconstruction_output, target_dir_output, image_gallery],
         )
         input_images.change(
             fn=update_gallery_on_upload,
-            inputs=[input_video, input_images],
+            inputs=[input_video, input_images, session_state],
             outputs=[reconstruction_output, target_dir_output, image_gallery],
         )
         demo.queue().launch(show_error=True, share=True)
-
-        # We thank VGGT for their excellent gradio implementation