Object Detection API Bounding box results are shifted

Software Python
1

Hi everyone,
I use internal inference engine to with yolo11 onnx file . Somehow detected object’s bounding box are shifted. When ball in ground is shift to upper points , when ball is high from ground it shifts lower points.

IMG_2418
IMG_24181512×2016 1.18 MB

IMG_2419
IMG_24192016×1512 1.13 MB

IMG_2420
IMG_24203024×4032 3.82 MB

These are the codes that I used. What can be the reason of this.

2

Hello @hatoktas,

Thank you for reaching out to us!

Can you confirm which ZED SDK version you are using ? If your are not using the latest (5.0.5), can you update your ZED SDK and confirm if the issue is still present? It looks like something that we’ve fixed in 5.0.3 and that we don’t see anymore.

If using the ZED SDK 5.0.5 does not solve the issue, can you confirm that your YOLOv11 model works properly by displaying its prediction on the same image as the one you shared with us?

Best,
Julien