-1

Is there any image annotation tool where I can import the bounding box coordinates (in PASCAL/VOC/YOLO format) and correct or edit the wrong annotations ? Wrong annotations are either wrong labels or shift in coordinates from actual object.

I have a huge video and image database for computer vision applications prepared by different people in the past. And there are lot of wrong annotations which needs to be corrected before I can use them.

Also, I have tried few annotation tools like microsoft vott and labelme.

Padfoot
  • 153
  • 1
  • 7
  • I voted to close because "Questions asking us to recommend or find a book, tool, software library, tutorial or other off-site resource are off-topic for Stack Overflow as they tend to attract opinionated answers and spam." This isn't a bad question in general, but it is just more appropriate for another forum where a discussion can be had. With that said, have you checked out the CVAT---computer vision annotation tool---which is maintained along with OpenCV? https://github.com/opencv/cvat – alkasm Jan 17 '20 at 08:00
  • @alkasm, I checked CVAT, but didn't found any option to import and edit the previous annotations. Thanks ! – Padfoot Jan 17 '20 at 08:17

1 Answers1

-1

Have you seen https://github.com/xiaqunfeng/BBox-Label-Tool? I always prefer this tool as the four coordinates are saved as a text file with the same name as the corresponding image. And later you can merge all the text files into one csv file using glob library or pandas library. In addition you can use Pycharm IDE to edit the csv and add the folder path behind each image coordinates as required by Yolo at once. After annotating through this tool, it takes me 5 minutes to compile a train.txt file that is ready to input into the model.

Areeb Muzaffar
  • 197
  • 2
  • 15