ss

app.py

@@ -40,11 +40,10 @@ example_inputs = [
     "examples/1.mp4",
     "examples/7.mp4",
     "examples/8.mp4",
-    "examples/36.mp4",
     "examples/39.mp4",
     "examples/10.mp4",
     "examples/30.mp4",
-    "examples/3.mp4",
+    
     "examples/35.mp4",
     "examples/40.mp4",
     "examples/2.mp4",
@@ -67,6 +66,7 @@ example_inputs = [
     "examples/input_20251202_031811.mp4",
     "examples/input_20251202_032007.mp4",
     "examples/teaser_1.mp4",
+    "examples/3.mp4",
     "examples/teaser_3.mp4",
     "examples/teaser_7.mp4",
     "examples/teaser_25.mp4",
@@ -442,7 +442,7 @@ with gr.Blocks(css=css, title="DKT", head=head_html) as demo:
             output_point_map0, output_point_map1, output_point_map2, output_point_map3
         ], 
         fn=on_example_submit,
-        examples_per_page=12,
+        examples_per_page=36,
         cache_examples=False
     )
 

dkt/pipelines/pipeline.py

@@ -29,7 +29,10 @@ from ..lora import GeneralLoRALoader
 
 from loguru import logger
 
+
+
 import spaces
+    
 
 class BasePipeline(torch.nn.Module):
 
@@ -976,23 +979,13 @@ class DKTPipeline:
     def moge_infer(self, input_image):        
         
         device = torch.device("cuda")
-
         
         self.moge_pipe = self.moge_pipe.to(device)
         self.moge_pipe.eval()
 
-
         # 5. 放到 GPU
         input_image = input_image.to(device=device, dtype=torch.float32)
 
-
-
-
-        # 🔴 必须补 batch 维度
-        if input_image.dim() == 3:
-            input_image = input_image.unsqueeze(0)  # (1, 3, H, W)
-        
-
         model_device = next(self.moge_pipe.parameters()).device
         print(f'input_image device: {input_image.device}, moge_pipe device: {model_device}, input shape: {input_image.shape}, input dtype: {input_image.dtype}')
         

examples/18.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8a384687c87416c80e20534d3889a83b3c523235fbea9631df22312e8450dfa0
-size 553261
+oid sha256:38a34cc7a7bd060e1cb891d35457b37e2ad91e8fac457273367500b65a8e1eb8
+size 1091805

examples/27.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b5e809c6ff324197946ac6c4c50d8b8d3dc05ab13da2665ef1b2e29186e6e008
-size 389973
+oid sha256:c6959f0dc0f5fde449cab87a0db04f2a14c23bc3df414b9e0474bd45bf901fbb
+size 893079

examples/28.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a58944143ce8d633f3d03eaf7caa28d61b40f50ad27a51bf6d36551c470a865b
-size 250182
+oid sha256:bd389e1fa12d73f8ded9181f9aa2b2586f8e150e855e96bf1df3c6420062bd12
+size 605351

examples/5eaeaff52b23787a3dc3c610655a49d2.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:48bbb2e58d64b6ae6d76961c754c06a043f4b412ef7eb0fad043d08860c59fed
-size 630823
+oid sha256:9f34029851307dea7724214a5ce5f9e0cc1f5283618d535dfd927ae4d5bc936e
+size 1371356

examples/73fc0b2a3af3474de27c7da0bfbf5faa.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8046f099a965a32b7ed825ac34c3f5cdd808395d28c75ac25870a976c2bc9a54
-size 900336
+oid sha256:c7619a66b098c36cf098e0d613b1132a4bc7a649bf32e0558574c6045dbf833b
+size 1837604

examples/9f2909760aff526070f169620ff38290.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3628fc6517b51adac1c61b295bb44823de3c834e6a522d03ae624c71706ffaaf
-size 898934
+oid sha256:c0f40eaaeefd6d4dff72cb944708e0e830ced4c3f97b4e8fb7ff7ce8b3f59c83
+size 2111360

examples/IMG_5703.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd7701a09bb7e874a081a04e5d06dd2a4f06bfd0849e91c617d3623bdb8069f1
+size 1897742

examples/episode_48-camera_third_view.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:46f8f64fc44c209fd18b0e521aae5dbc790a03b1d9fa6a40852bcffb51971357
-size 2187220
+oid sha256:59efd48910999b4ed0c356438c676647965c9a2e4dde78e70041ed150fab3e57
+size 776108

examples/extra_5.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:45da13055895736f08510abcdd716087d8065deb7e3be51105f0011d1e4af427
-size 549538
+oid sha256:87105db3c75e0579118a52798a27d0373810dc2fb0690be6126d5e1d18ab9ed1
+size 1088857

examples/extra_9.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0730c76dc4395c67507922e69ace6db3bd809dbb032a34cb8bc7d1d4ecda6566
-size 239514
+oid sha256:67ba8aef9ac1d4370f70896399a7dc093b0bc9aff9cd4e7e504dbe7eaf3d3016
+size 547381

examples/input_20251128_121408.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e2daa715eb4bd6652d0baab4ac8a23efe62ecb2589054f5b4e6ab7fcbdbe45cf
-size 1488311
+oid sha256:b23bfaf76a6d3a9bcb8124c6f9ace836885e4fafcad110cc7f9095752f96b324
+size 2722066

examples/input_20251128_122722.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6e97aef540e9c04a06d0b1ab88c8a9263641ddcbe4d5cdb2286059e4a20ca503
-size 314882
+oid sha256:5531cbf5890dfca572fcc1eb28a859f4cd09ae2e822f9bb579fe80eeca2d6962
+size 692739

examples/input_20251202_031811.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ce35b4b96e4f92e9c554774691e2283b7bd4f122e26aefb83057fc6be3a5aaac
-size 280854
+oid sha256:5530e4322fefacb7d75245fcc13ec64085181e66c546438cfc935fe5ba5fecd2
+size 545173

examples/input_20251202_032007.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:55f9570bcf6eff0dd5ae3fd7e9212f7c44b1c81900c5843b37b71be081eb8e5f
-size 252242
+oid sha256:cd9cdf1fd5ee8820d74882405c88cb16860c2c1a1d07662538390a175b886080
+size 561135

examples/teaser_1.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:34537e88f30114e06e4eabe5880eb2a54f0202672e33a8e94c385a83d2113a52
-size 471670
+oid sha256:4f454ee9e5d7d35ac17ce515c05c76feca434d38d0fc3c674aaf4e28ced95869
+size 881391

examples/teaser_25.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ce3ef29c9b359b2929f4b3c210adb182bac8ec9b349d7e7fc3eb976a41380af2
-size 1769586
+oid sha256:5115a950b878631f8aefb00bd7f39b6f73ed3a636976df462b190d34df41ade4
+size 1087457

examples/teaser_3.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ed81db20b9a786065c21546f1deac2d1953b05f4211e92cd409c0e548d1642cf
-size 901138
+oid sha256:52c3f306ee1ac2db9c7284908e2798d41e7c47d6796fdb17c7b429a8e37298bb
+size 2713463

examples/teaser_7.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9ca62bfb7c9e2745cd8b61e0d47f8ef66c7c66520df429ae6b6795c320f11809
-size 462918
+oid sha256:59bd55c5dfcf8d362af1c0942c0dd5c3d11c4c8b3e3832b390275c4f3cb6f44c
+size 1021290

tools/common_utils.py

@@ -12,6 +12,12 @@ def save_video(frames, save_path, fps, quality=9, ffmpeg_params=None):
         frames[0].save(save_path.replace('.mp4', '.png'))
         return
 
+    # Ensure browser-compatible pixel format (yuv420p) to avoid Gradio conversion warning
+    if ffmpeg_params is None:
+        ffmpeg_params = ['-pix_fmt', 'yuv420p']
+    elif '-pix_fmt' not in ffmpeg_params:
+        ffmpeg_params = list(ffmpeg_params) + ['-pix_fmt', 'yuv420p']
+    
     writer = imageio.get_writer(save_path, fps=fps, quality=quality, ffmpeg_params=ffmpeg_params)
     for frame in tqdm(frames, desc="Saving video"):
         frame = np.array(frame)