0

Im trying to use CVAT in order to tag a dataset that I have. I'm trying to label the gender of pedestrians, etc, and I have a yolov4 network that actually marks all pedestrians with bboxes. I thought I could save me some trouble by importing the output of yolo (a file that describes all bounding boxes for all the images) to CVAT and only tag the already existing bboxes. I have tried the option of 'importing datasets' to my CVAT project, following this guide: https://opencv.github.io/cvat/docs/manual/advanced/formats/format-yolo/#yolo-export (Format of yolo1.1) But when uploading the archive zip file I get this error: Request failed with status code 500. "KeyError: 0\n"..

Perhaps someone knows how to handle this? This is so frustrating :( Thanks, Shani

I have tried the guide mentioned above

shani zehavi
  • 1

1 Answers1

0

I do not know if we have the same problem, but like you, I tried to upload a previously annotated (tagged) file to CVAT and got the following error.

Error: Request failed with status code 500. "KeyError: 'points'\n"

So, I opened the file and found that inside the annotated file, the bounding boxes are referred to as 'polygon frame' instead of 'box frame'. Once I renamed polygon to box, it worked. I am using CVAT1.1 format, BTW.

Now your use case might be different. I would suggest you to debug this situation by doing a manual annotation first. Then export the file in Yolo format. Then open the exported file and the file you are trying to import side by side. Then try to compare what's the difference and try to bring changes accordingly.

Edit: Perhaps these two links can help?

https://github.com/opencv/cvat/issues/5654

https://github.com/opencv/cvat/issues/5653

Muhammad Raihan
  • 11
  • 3