Spaces:
Running
on
Zero
Running
on
Zero
Have you finetuned the model?
#5
by
wytalfred
- opened
Why I can't get correct detection boxes on original Qwen/Qwen2-VL-7B-Instruct model? The output format is right but the numbers are very wrong.
Looking forward to your help, thank you in advance!
Hi,
when I remember correctly. The bounding boxes are normalized by 1000. So you've to recalculate them to the original size
Hi,
when I remember correctly. The bounding boxes are normalized by 1000. So you've to recalculate them to the original size
It worked! Thank you! I wonder where do you get this information? I don't see it anywhere.