How to train yolov7-mask？

WongKinYiu / yolov7

Implementation of paper - YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors

GNU General Public License v3.0

13.39k stars 4.23k forks source link

How to train yolov7-mask？ #669

Open Ohandsomeboy opened 2 years ago

alaap001 commented 2 years ago

+1 I know that the training code is not yet released so far based on previous threads, but is there any UPDATE on that as to when we will release the train code? @WongKinYiu

WongKinYiu commented 2 years ago

Currently we release training code in u7 branch.

alaap001 commented 2 years ago

That is nice. I see it was recently pushed in that branch. Thankyou for the release. I have 1 query tho, Training format is still unclear. What is the format of labels?

Like in OD we have img1.jpg and img1.txt (xc,yc,w,h)

What about seg? Is it using the normal coco_annotation_train.json file? Or do we need .txt files with some specific format?

WongKinYiu commented 2 years ago

Polygon: x0 y0 x1 y1 ... xn yn

alaap001 commented 2 years ago

Oh ok. Understood. Thanks. Also, Is there a script to convert from coco to this format? Also, can I put negative samples in this? like empty labels txts.

WongKinYiu commented 2 years ago

JSON2YOLO

iwaitu commented 2 years ago

@WongKinYiu Where is the python segment/train.py config file named yolov7-seg.yaml ?