Update app.py

app.py

@@ -51,163 +51,173 @@ print("LD_LIBRARY_PATH:", os.environ['LD_LIBRARY_PATH'])
 
 from stf_utils import STFPipeline
 
+stf_pipeline = STFPipeline()
 
-def partial_fields(target_class, kwargs):
-    return target_class(**{k: v for k, v in kwargs.items() if hasattr(target_class, k)})
+@spaces.GPU(duration=120)
+def gpu_wrapped_stf_pipeline_execute(audio_path):
+    return stf_pipeline.execute(audio_path)
 
-# set tyro theme
-tyro.extras.set_accent_color("bright_cyan")
-args = tyro.cli(ArgumentConfig)
+    
 
-# specify configs for inference
-inference_cfg = partial_fields(InferenceConfig, args.__dict__)  # use attribute of args to initial InferenceConfig
-crop_cfg = partial_fields(CropConfig, args.__dict__)  # use attribute of args to initial CropConfig
 
-# gradio_pipeline = GradioPipeline(
-#     inference_cfg=inference_cfg,
-#     crop_cfg=crop_cfg,
-#     args=args
-# )
 
-@spaces.GPU(duration=120)
-def gpu_wrapped_execute_video(*args, **kwargs):
-    return gradio_pipeline.execute_video(*args, **kwargs)
 
-@spaces.GPU(duration=120)
-def gpu_wrapped_execute_image(*args, **kwargs):
-    return gradio_pipeline.execute_image(*args, **kwargs)
+# def partial_fields(target_class, kwargs):
+#     return target_class(**{k: v for k, v in kwargs.items() if hasattr(target_class, k)})
 
-@spaces.GPU(duration=120)
-def gpu_wrapped_stf_pipeline_execute(audio_path):
-    return stf_pipeline.execute(audio_path)
+# # set tyro theme
+# tyro.extras.set_accent_color("bright_cyan")
+# args = tyro.cli(ArgumentConfig)
 
-@spaces.GPU(duration=120)
-def gpu_wrapped_elevenlabs_pipeline_generate_voice(text, voice):
-    return elevenlabs_pipeline.generate_voice(text, voice)
+# # specify configs for inference
+# inference_cfg = partial_fields(InferenceConfig, args.__dict__)  # use attribute of args to initial InferenceConfig
+# crop_cfg = partial_fields(CropConfig, args.__dict__)  # use attribute of args to initial CropConfig
+
+# # gradio_pipeline = GradioPipeline(
+# #     inference_cfg=inference_cfg,
+# #     crop_cfg=crop_cfg,
+# #     args=args
+# # )
+
+# @spaces.GPU(duration=120)
+# def gpu_wrapped_execute_video(*args, **kwargs):
+#     return gradio_pipeline.execute_video(*args, **kwargs)
+
+# @spaces.GPU(duration=120)
+# def gpu_wrapped_execute_image(*args, **kwargs):
+#     return gradio_pipeline.execute_image(*args, **kwargs)
+
+# @spaces.GPU(duration=120)
+# def gpu_wrapped_stf_pipeline_execute(audio_path):
+#     return stf_pipeline.execute(audio_path)
+
+# @spaces.GPU(duration=120)
+# def gpu_wrapped_elevenlabs_pipeline_generate_voice(text, voice):
+#     return elevenlabs_pipeline.generate_voice(text, voice)
 
     
-def is_square_video(video_path):
-    video = cv2.VideoCapture(video_path)
+# def is_square_video(video_path):
+#     video = cv2.VideoCapture(video_path)
 
-    width = int(video.get(cv2.CAP_PROP_FRAME_WIDTH))
-    height = int(video.get(cv2.CAP_PROP_FRAME_HEIGHT))
+#     width = int(video.get(cv2.CAP_PROP_FRAME_WIDTH))
+#     height = int(video.get(cv2.CAP_PROP_FRAME_HEIGHT))
 
-    video.release()
-    if width != height:
-        raise gr.Error("Error: the video does not have a square aspect ratio. We currently only support square videos")
+#     video.release()
+#     if width != height:
+#         raise gr.Error("Error: the video does not have a square aspect ratio. We currently only support square videos")
 
-    return gr.update(visible=True)
+#     return gr.update(visible=True)
     
 
-# assets
-title_md = "assets/gradio_title.md"
-example_portrait_dir = "assets/examples/source"
-example_video_dir = "assets/examples/driving"
-data_examples = [
-    [osp.join(example_portrait_dir, "s9.jpg"), osp.join(example_video_dir, "d0.mp4"), True, True, True, True],
-    [osp.join(example_portrait_dir, "s6.jpg"), osp.join(example_video_dir, "d0.mp4"), True, True, True, True],
-    [osp.join(example_portrait_dir, "s10.jpg"), osp.join(example_video_dir, "d0.mp4"), True, True, True, True],
-    [osp.join(example_portrait_dir, "s5.jpg"), osp.join(example_video_dir, "d18.mp4"), True, True, True, True],
-    [osp.join(example_portrait_dir, "s7.jpg"), osp.join(example_video_dir, "d19.mp4"), True, True, True, True],
-    [osp.join(example_portrait_dir, "s22.jpg"), osp.join(example_video_dir, "d0.mp4"), True, True, True, True],
-]
-#################### interface logic ####################
-
-
-
-# Define components first
-eye_retargeting_slider = gr.Slider(minimum=0, maximum=0.8, step=0.01, label="target eyes-open ratio")
-lip_retargeting_slider = gr.Slider(minimum=0, maximum=0.8, step=0.01, label="target lip-open ratio")
-retargeting_input_image = gr.Image(type="filepath")
-output_image = gr.Image(type="numpy")
-output_image_paste_back = gr.Image(type="numpy")
-output_video = gr.Video()
-output_video_concat = gr.Video()
+# # assets
+# title_md = "assets/gradio_title.md"
+# example_portrait_dir = "assets/examples/source"
+# example_video_dir = "assets/examples/driving"
+# data_examples = [
+#     [osp.join(example_portrait_dir, "s9.jpg"), osp.join(example_video_dir, "d0.mp4"), True, True, True, True],
+#     [osp.join(example_portrait_dir, "s6.jpg"), osp.join(example_video_dir, "d0.mp4"), True, True, True, True],
+#     [osp.join(example_portrait_dir, "s10.jpg"), osp.join(example_video_dir, "d0.mp4"), True, True, True, True],
+#     [osp.join(example_portrait_dir, "s5.jpg"), osp.join(example_video_dir, "d18.mp4"), True, True, True, True],
+#     [osp.join(example_portrait_dir, "s7.jpg"), osp.join(example_video_dir, "d19.mp4"), True, True, True, True],
+#     [osp.join(example_portrait_dir, "s22.jpg"), osp.join(example_video_dir, "d0.mp4"), True, True, True, True],
+# ]
+# #################### interface logic ####################
+
+
+
+# # Define components first
+# eye_retargeting_slider = gr.Slider(minimum=0, maximum=0.8, step=0.01, label="target eyes-open ratio")
+# lip_retargeting_slider = gr.Slider(minimum=0, maximum=0.8, step=0.01, label="target lip-open ratio")
+# retargeting_input_image = gr.Image(type="filepath")
+# output_image = gr.Image(type="numpy")
+# output_image_paste_back = gr.Image(type="numpy")
+# output_video = gr.Video()
+# output_video_concat = gr.Video()
 
         
 
 
-def run_end_to_end(image_path, text, voice, input_video, flag_relative, flag_do_crop, flag_remap, flag_crop_driving_video, male): #, animal):
+# def run_end_to_end(image_path, text, voice, input_video, flag_relative, flag_do_crop, flag_remap, flag_crop_driving_video, male): #, animal):
     
-    # # animal 체크 여부에 따라 다른 pipeline 사용
-    # if animal:
-    #     gradio_pipeline = GradioPipelineAnimal(
-    #         inference_cfg=inference_cfg,
-    #         crop_cfg=crop_cfg,
-    #         args=args
-    #     )
-    # else:
-    #     gradio_pipeline = GradioPipeline(
-    #         inference_cfg=inference_cfg,
-    #         crop_cfg=crop_cfg,
-    #         args=args
-    #     )
-
-
-    if input_video is None:
+#     # # animal 체크 여부에 따라 다른 pipeline 사용
+#     # if animal:
+#     #     gradio_pipeline = GradioPipelineAnimal(
+#     #         inference_cfg=inference_cfg,
+#     #         crop_cfg=crop_cfg,
+#     #         args=args
+#     #     )
+#     # else:
+#     #     gradio_pipeline = GradioPipeline(
+#     #         inference_cfg=inference_cfg,
+#     #         crop_cfg=crop_cfg,
+#     #         args=args
+#     #     )
+
+
+#     if input_video is None:
     
-        if not male:
-            stf_pipeline = STFPipeline()
-        else:
-            stf_pipeline = STFPipeline(template_video_path="/home/user/app/stf/TEMP/Cam2_2309071202_0012_Natural_Looped.mp4",
-               config_path="/home/user/app/stf/TEMP/front_config_v3.json",
-               checkpoint_path="/home/user/app/stf/TEMP/0157.pth",
-               )
-
-    if input_video is None:
-        #audio_path = elevenlabs_pipeline.generate_voice(text, voice)
-        audio_path = gpu_wrapped_elevenlabs_pipeline_generate_voice(text, voice)
-        #driving_video_path = stf_pipeline.execute(audio_path)
-        driving_video_path = gpu_wrapped_stf_pipeline_execute(audio_path)
-    else:
-        driving_video_path = input_video
-        os.makedirs("animations",exist_ok=True)
-        audio_path = osp.join("animations", Path(input_video).stem+".wav")
-        extract_audio(driving_video_path, audio_path)
-
-
-    #output_path, crop_output_path = gradio_pipeline.execute_video(
-    output_path, crop_output_path = gpu_wrapped_execute_video(
-            input_image_path=image_path,
-            input_video_path=driving_video_path,
-            # input_driving_video_pickle_path=None,
-            flag_do_crop_input=flag_do_crop,
-            flag_remap_input=flag_remap,
-            flag_relative_input=flag_relative,
-            # driving_multiplier=1.0,
-            # flag_stitching=False,
-            # flag_crop_driving_video_input=flag_crop_driving_video,
-            # scale=2.3,
-            # vx_ratio=0.0,
-            # vy_ratio=-0.125,
-            # scale_crop_driving_video=2.2,
-            # vx_ratio_crop_driving_video=0.0,
-            # vy_ratio_crop_driving_video=-0.1,
-            # tab_selection=None,
-            audio_path=audio_path
-            )
+#         if not male:
+#             stf_pipeline = STFPipeline()
+#         else:
+#             stf_pipeline = STFPipeline(template_video_path="/home/user/app/stf/TEMP/Cam2_2309071202_0012_Natural_Looped.mp4",
+#                config_path="/home/user/app/stf/TEMP/front_config_v3.json",
+#                checkpoint_path="/home/user/app/stf/TEMP/0157.pth",
+#                )
+
+#     if input_video is None:
+#         #audio_path = elevenlabs_pipeline.generate_voice(text, voice)
+#         audio_path = gpu_wrapped_elevenlabs_pipeline_generate_voice(text, voice)
+#         #driving_video_path = stf_pipeline.execute(audio_path)
+#         driving_video_path = gpu_wrapped_stf_pipeline_execute(audio_path)
+#     else:
+#         driving_video_path = input_video
+#         os.makedirs("animations",exist_ok=True)
+#         audio_path = osp.join("animations", Path(input_video).stem+".wav")
+#         extract_audio(driving_video_path, audio_path)
+
+
+#     #output_path, crop_output_path = gradio_pipeline.execute_video(
+#     output_path, crop_output_path = gpu_wrapped_execute_video(
+#             input_image_path=image_path,
+#             input_video_path=driving_video_path,
+#             # input_driving_video_pickle_path=None,
+#             flag_do_crop_input=flag_do_crop,
+#             flag_remap_input=flag_remap,
+#             flag_relative_input=flag_relative,
+#             # driving_multiplier=1.0,
+#             # flag_stitching=False,
+#             # flag_crop_driving_video_input=flag_crop_driving_video,
+#             # scale=2.3,
+#             # vx_ratio=0.0,
+#             # vy_ratio=-0.125,
+#             # scale_crop_driving_video=2.2,
+#             # vx_ratio_crop_driving_video=0.0,
+#             # vy_ratio_crop_driving_video=-0.1,
+#             # tab_selection=None,
+#             audio_path=audio_path
+#             )
    
-    return output_path, crop_output_path
+#     return output_path, crop_output_path
 
 
 
 
-###### 테스트중 ######
+# ###### 테스트중 ######
     
 
-stf_pipeline = STFPipeline()
+# stf_pipeline = STFPipeline()
+
+# with gr.Blocks(theme=gr.themes.Soft()) as demo:
+#     stf_button = gr.Button("stf test", variant="primary")
+#     stf_button.click(
+#                 fn=gpu_wrapped_stf_pipeline_execute,
+#                 inputs=[
+#                     audio_path
+#                 ],
+#                 outputs=[driving_video_path]
+#             )
 
-with gr.Blocks(theme=gr.themes.Soft()) as demo:
-    stf_button = gr.Button("stf test", variant="primary")
-    stf_button.click(
-                fn=gpu_wrapped_stf_pipeline_execute,
-                inputs=[
-                    audio_path
-                ],
-                outputs=[driving_video_path]
-            )
-
-###### 테스트중 ######
+# ###### 테스트중 ######
 
 
 # with gr.Blocks(theme=gr.themes.Soft()) as demo: