CIHP-PGN : Crowd Instance-level Human Parsing via Part Grouping Network
저장경로를 datasets/silhouatte/으로 하며 파일에 _sil.png로 한다.
###### gryscale silhouatte
sil_dir = os.path.join(args.output, 'silhouette')
if not os.path.exists(sil_dir):
os.makedirs(sil_dir)
# Iterate over training steps.
...
###### gryscale silhouatte
gray = cv2.cvtColor(msk[0], cv2.COLOR_BGR2GRAY)
(thresh, blackAndWhiteImage) = cv2.threshold(gray, 1, 255, cv2.THRESH_BINARY)
cv2.imwrite('{}/{}_sil.png'.format(sil_dir, img_id), blackAndWhiteImage)
source : datasets/images/bill.jpg |
output:output/cihp_edge_maps/bill.png edge추출(사용x) |
output:cihp_parsing_maps 사용x |
|
output:output/silhouatte/bill_sil.png 실루엣 추출-detectron2에서 사용 |
참고자료
https://github.com/Engineering-Course/CIHP_PGN
'AI > pose with style' 카테고리의 다른 글
YOLO V7 - installation (0) | 2023.02.16 |
---|---|
Detection Human and garments (0) | 2023.02.15 |
Pose with Style - Test (0) | 2023.02.14 |
CIHP-PGN Multi-Person (0) | 2023.02.14 |
detectron2 (0) | 2023.02.14 |
댓글