update

app.py

@@ -15,22 +15,26 @@ from datetime import datetime
 from vggsfm_code.hf_demo import demo_fn
 from omegaconf import DictConfig, OmegaConf
 from viz_utils.viz_fn import add_camera
-
+import glob
 # 
 from scipy.spatial.transform import Rotation
 import PIL
 
 
-import spaces
+# import spaces
 
-@spaces.GPU
+# @spaces.GPU
 def vggsfm_demo(
-    input_image,
     input_video,
+    input_image,
     query_frame_num,
-    max_query_pts
-    # grid_size: int = 10,
+    max_query_pts=4096,
 ):
+
+    if input_video is not None:            
+        if not isinstance(input_video, str):
+            input_video = input_video["video"]["path"]
+    
     cfg_file = "vggsfm_code/cfgs/demo.yaml"
     cfg = OmegaConf.load(cfg_file)
 
@@ -61,6 +65,7 @@ def vggsfm_demo(
 
         fps = vs.get(cv2.CAP_PROP_FPS)
 
+
         frame_rate = 1
         frame_interval = int(fps * frame_rate)
         
@@ -70,12 +75,14 @@ def vggsfm_demo(
         while video_frame_num<=max_input_image:
             (gotit, frame) = vs.read()
             count +=1
+
+            if not gotit:
+                break
             
             if count % frame_interval == 0:
                 cv2.imwrite(target_dir_images+"/"+f"{video_frame_num:06}.png", frame)
                 video_frame_num+=1
-            if not gotit:
-                break
+                
         if video_frame_num<3:
             return None, "Please input at least three frames"
     else:
@@ -86,13 +93,16 @@ def vggsfm_demo(
     print(f"Files have been copied to {target_dir_images}")
     cfg.SCENE_DIR = target_dir
     
-    try:
-        predictions = demo_fn(cfg)
-    except:
-        return None, "Something seems to be incorrect. Please verify that your inputs are formatted correctly. If the issue persists, kindly create a GitHub issue for further assistance."
+    # try:
+    predictions = demo_fn(cfg)
+    # except:
+    # return None, "Something seems to be incorrect. Please verify that your inputs are formatted correctly. If the issue persists, kindly create a GitHub issue for further assistance."
     
-    glbfile = vggsfm_predictions_to_glb(predictions)
+    glbscene = vggsfm_predictions_to_glb(predictions)
     
+    glbfile = target_dir + "/glbscene.glb"
+    glbscene.export(file_obj=glbfile)    
+
     
     print(input_image)
     print(input_video)
@@ -132,12 +142,6 @@ def vggsfm_predictions_to_glb(predictions):
         cur_cam_color = camera_edge_colors[idx % len(camera_edge_colors)]
         cur_focal = intrinsics_opencv[idx, 0, 0]
 
-        # cur_image_path = raw_image_paths[idx]
-
-        # cur_image = np.array(PIL.Image.open(cur_image_path))
-        # add_camera(glbscene, cam_to_world, cur_cam_color, image=None, imsize=cur_image.shape[1::-1], 
-                #    focal=None,screen_width=0.3)
-
         add_camera(glbscene, cam_to_world, cur_cam_color, image=None, imsize=(1024,1024), 
                    focal=None,screen_width=0.35)
 
@@ -150,35 +154,59 @@ def vggsfm_predictions_to_glb(predictions):
     rot[:3, :3] = Rotation.from_euler('y', np.deg2rad(180)).as_matrix()
     glbscene.apply_transform(np.linalg.inv(np.linalg.inv(extrinsics_opencv_4x4[0]) @ opengl_mat @ rot))
 
-    glbfile = "glbscene.glb"
-    glbscene.export(file_obj=glbfile)    
-    return glbfile
+    # glbfile = "glbscene.glb"
+    # glbscene.export(file_obj=glbfile)    
+    return glbscene
+
+# apple_video = "vggsfm_code/examples/videos/apple_video.mp4"
+# os.path.join(os.path.dirname(__file__), "apple_video.mp4")
+british_museum_video = "vggsfm_code/examples/videos/british_museum_video.mp4"
+
+# os.path.join(os.path.dirname(__file__), "british_museum_video.mp4")
+cake_video = "vggsfm_code/examples/videos/cake_video.mp4"
 
+# os.path.join(os.path.dirname(__file__), "cake_video.mp4")
 
 
 
+# apple_images = glob.glob(f'vggsfm_code/examples/apple/images/*')
+cake_images = glob.glob(f'vggsfm_code/examples/cake/images/*')
+british_museum_images = glob.glob(f'vggsfm_code/examples/british_museum/images/*')
 
 if True:
     demo = gr.Interface(
         title="🎨 VGGSfM: Visual Geometry Grounded Deep Structure From Motion",
-        description="<div style='text-align: left;'> \
-        <p>Welcome to <a href='https://github.com/facebookresearch/vggsfm' target='_blank'>VGGSfM</a>!",
         fn=vggsfm_demo,
         inputs=[
-            gr.File(file_count="multiple", label="Input Images", interactive=True),
             gr.Video(label="Input video", interactive=True),
+            gr.File(file_count="multiple", label="Input Images", interactive=True),
             gr.Slider(minimum=1, maximum=10, step=1, value=5, label="Number of query images"),
             gr.Slider(minimum=512, maximum=4096, step=1, value=1024, label="Number of query points"),
         ],
         outputs=[gr.Model3D(label="Reconstruction"), gr.Textbox(label="Log")],
+        examples=[
+            # [apple_video, apple_images, 5, 2048],
+            [cake_video, cake_images, 3, 4096],
+            [british_museum_video, british_museum_images, 2, 4096],],
         cache_examples=True,
         # allow_flagging=False,
         allow_flagging='never',  # Updated from False to 'never'
-        concurrency_limit=1  # Added concurrency_limit to Interface
+        concurrency_limit=1,  # Added concurrency_limit to Interface
+        description = """<div style="text-align: left;"> 
+        <p>Welcome to <a href="https://github.com/facebookresearch/vggsfm" target="_blank">VGGSfM</a> demo! 
+        This space demonstrates 3D reconstruction from input image frames. </p> 
+        <p>To get started quickly, you can click on our examples. If you want to reconstruct your own data, simply: </p> 
+        <ul style="display: inline-block; text-align: left;"> 
+            <li>upload the images (.jpg, .png, etc.), or </li> 
+            <li>upload a video (.mp4, .mov, etc.) </li> 
+        </ul> 
+        <p>If both images and videos are uploaded, the demo will only reconstruct the uploaded images. By default, we extract one image frame per second from the input video. To prevent crashes on the Hugging Face space, we currently limit reconstruction to the first 20 image frames. </p> 
+        <p>For more details, check our <a href="https://github.com/facebookresearch/vggsfm" target="_blank">GitHub Repo</a> ⭐</p> 
+        <p>(Please note that running reconstruction on Hugging Face space is slower than on a local machine.) </p> 
+        </div>""",
     )
-    demo.queue(max_size=20).launch(debug=True)
 
-    # demo.launch(debug=True, share=True)
+    demo.launch(debug=True, share=True)
 else:
     import glob
     files = glob.glob(f'vggsfm_code/examples/cake/images/*', recursive=True)

images_to_videos.py

@@ -0,0 +1,28 @@
+import cv2
+import os
+
+# Parameters
+
+name = "cake"
+folder_path = f'vggsfm_code/examples/{name}/images'  # Update with the path to your images
+video_path = f'{name}_video.mp4'
+fps = 1  # frames per second
+
+# Get all image files from the directory
+images = [img for img in os.listdir(folder_path)]
+images.sort()  # Sort the images by name
+
+# Read the first image to get the size
+frame = cv2.imread(os.path.join(folder_path, images[0]))
+height, width, layers = frame.shape
+
+# Define the codec and create VideoWriter object
+fourcc = cv2.VideoWriter_fourcc(*'mp4v')  # or 'x264' for codec
+video = cv2.VideoWriter(video_path, fourcc, fps, (width, height))
+
+# Add images to video
+for image in images:
+    video.write(cv2.imread(os.path.join(folder_path, image)))
+
+# Release the video writer
+video.release()