0

I labeled images with bounding box and then used yolov2 (python 3.7) for object detection and It was successful (https://github.com/deep-diver/Soccer-Ball-Detection-YOLOv2/blob/master/YOLOv2-Train.ipynb). I need to use yolov2 for semantic segmentation using python. Which image-labeling software can I use for semantic segmentation which its output is compatible with yolo? How can I modify the code in the aforementioned website to use yolov2 for this matter? Or do we have any example for that?

Thank you for your help in advance.

Abbas,

Abbas
  • 71
  • 1
  • 8

2 Answers2

0

None of the YOLO algorithms perform Semantic Segmentation. They are Object Detection algorithms. In your case, you can use any of the following algorithms for segmentation: DeepLab, UNet, PSPNet, EfficientNet, etc.

planet_pluto
  • 742
  • 4
  • 15
  • I found "Mask R-CNN" for semantic segmentation. Any idea bout this method? Thanks, – Abbas Oct 21 '20 at 19:53
  • Remember, Mask RCNN and YOLACT/YOLACT++ are instance segmentation models and not semantic segmentation. Both the architectures are quite complex, especially the Mask RCNN. If you're starting in this field, I would suggest you to look at the models I had mentioned in my post. – planet_pluto Oct 22 '20 at 05:39
0

Hello You can use YOLACT/YOLACT++ https://github.com/dbolya/yolact It uses same labelme annotations and you just have to edit the cofig.py a little and you will get the desired results. I am using it and it is superb.

Yolact is a kind of transformation of YOLO and it means You Look at the Coefficient.

Areeb Muzaffar
  • 197
  • 2
  • 15
  • Hi Areeb, Thank you for your help. I will use YOLACT for semantic segmentation. – Abbas Oct 21 '20 at 13:51
  • YOLACT is indeed a very good model but I wouldn't advise it for someone who's starting their journey in semantic segmentation. – planet_pluto Oct 21 '20 at 19:10