Issue with OWLv2's image-guided-detection mode.

[taherpat]

I have tried endless times to recreate the results from the tutorial notebook of https://github.com/NielsRogge/Transformers-Tutorials/blob/master/OWLv2/Zero_and_one_shot_object_detection_with_OWLv2.ipynb but I still cannot reproduce similar results.

I have also tried keeping the notebook exactly the same and just changing the source and target URLs. It still doesn't work as in the original notebook.

This works absolutely well when I am going through the original method and not using the transformer's inbuilt image_guided_detection.

Any suggestions would be really great.

Final Result with original method:

Final Results with Transformers methodL

As for the code, it is exactly the same notebook as in link only with the following target and source image links- target_url = 'http://images.cocodataset.org/train2017/000000558809.jpg' source_url = 'http://images.cocodataset.org/train2017/000000129908.jpg'

@NielsRogge