Mirrors/modelscope
1
0
Fork 0
mirror of https://github.com/modelscope/modelscope.git synced 2025-12-25 12:39:25 +01:00
Code Issues Packages Projects Releases Wiki Activity

adjust video_human_matting output of video to support demo service

Browse Source 
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/11486472
This commit is contained in:
jinmao.yk
2023-01-31 14:34:05 +00:00
committed by wenmeng.zwm
parent 3eda77aa4f
commit a272d00c54
2 changed files with 11 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
modelscope/outputs/outputs.py
View File

@@ -534,8 +534,9 @@ TASK_OUTPUTS = {
# video human matting result for a single video
# {
# "masks": [np.array # 2D array with shape [height, width]]
# "output_video": "path_to_matting_video"
# }
Tasks.video_human_matting: [OutputKeys.MASKS],
Tasks.video_human_matting: [OutputKeys.MASKS, OutputKeys.OUTPUT_VIDEO],
# ============ nlp tasks ===================

16
modelscope/pipelines/cv/video_human_matting_pipeline.py
View File

@@ -37,9 +37,11 @@ class VideoHumanMattingPipeline(Pipeline):
def preprocess(self, input) -> Input:
return input
def forward(self, input: Dict[str, Any]) -> Dict[str, Any]:
def forward(self, input: Dict[str, Any],
**forward_params) -> Dict[str, Any]:
video_path = input['video_input_path']
out_path = input['output_path']
render = forward_params.get('render', False)
video_input = cv2.VideoCapture(video_path)
fps = video_input.get(cv2.CAP_PROP_FPS)
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
@@ -58,19 +60,19 @@ class VideoHumanMattingPipeline(Pipeline):
frame_tensor = preprocess(frame)
pha, *rec = self.model.model(
frame_tensor.to(self.device), *rec, downsample_ratio=scale)
com = pha * 255
com = com.repeat(1, 3, 1, 1)
com = com[0].data.cpu().numpy().transpose(1, 2,
0).astype(np.uint8)
mask = pha * 255
mask = mask[0].data.cpu().numpy().transpose(1, 2, 0)
com = mask.repeat(3, 2).astype(np.uint8)
video_save.write(com)
masks.append(com / 255)
masks.append((mask / 255).astype(np.uint8))
success, frame = video_input.read()
logger.info('matting process done')
video_input.release()
video_save.release()
return {
OutputKeys.MASKS: masks,
OutputKeys.MASKS: None if render else masks,
OutputKeys.OUTPUT_VIDEO: out_path
}
def postprocess(self, inputs: Dict[str, Any]) -> Dict[str, Any]: