Update utils/video_process.py

utils/video_process.py

@@ -19,7 +19,7 @@ from models.av_mossformer2_tse.faceDetector.s3fd import S3FD
 
 from .decode import decode_one_audio_AV_MossFormer2_TSE_16K
 
-
+import spaces
 
 def process_tse(args, model, device, data_reader, output_wave_dir):
 	video_args = args_param()
@@ -162,6 +162,7 @@ def scene_detect(video_args):
 		sys.stderr.write('%s - scenes detected %d\n'%(video_args.videoFilePath, len(sceneList)))
 	return sceneList
 
+@spaces.GPU(duration=100)
 def inference_video(video_args):
 	# GPU: Face detection, output is the list contains the face location and score in this frame
 	DET = S3FD(device=video_args.device)
@@ -181,6 +182,7 @@ def inference_video(video_args):
 		pickle.dump(dets, fil)
 	return dets
 
+
 def bb_intersection_over_union(boxA, boxB, evalCol = False):
 	# CPU: IOU Function to calculate overlap between two image
 	xA = max(boxA[0], boxB[0])
@@ -268,6 +270,7 @@ def crop_video(video_args, track, cropFile):
 	return {'track':track, 'proc_track':dets}
 
 
+@spaces.GPU(duration=100)
 def evaluate_network(files, video_args, args):
 
 	est_sources = []