Inaccurate Inference Results When Using OWLv2's `image_guided_detection`

[Itto1992]

System Info

• transformers version: 4.35.2
• Platform: Linux-5.15.120+-x86_64-with-glibc2.35
• Python version: 3.10.12
• Huggingface_hub version: 0.19.4
• Safetensors version: 0.4.1
• Accelerate version: not installed
• Accelerate config: not found
• PyTorch version (GPU?): 2.1.0+cu118 (True)
• Tensorflow version (GPU?): 2.14.0 (True)
• Flax version (CPU?/GPU?/TPU?): 0.7.5 (gpu)
• Jax version: 0.4.20
• JaxLib version: 0.4.20
• Using GPU in script?: Yes
• Using distributed or parallel set-up in script?: No

Who can help?

@ArthurZucker
@younesbelkada
@amyeroberts

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

Run the sample code of OWLv2: https://huggingface.co/docs/transformers/model_doc/owlv2#transformers.Owlv2ForObjectDetection.image_guided_detection.example

Expected behavior

• The bounding boxes intended to encircle two cats in the sample code appear in inappropriate positions.
• In the sample code (https://huggingface.co/docs/transformers/model_doc/owlv2#transformers.Owlv2ForObjectDetection.image_guided_detection.example), 13 boxes appear in locations unrelated to cats. I ran the exact same code in Google Colab and was able to replicate the same inference results.
• The previous model (OWLViT) was able to detect without any issues.

I am unsure if this behavior is expected, so I would appreciate some advice. If you are aware of any methods to improve the inference performance, please let me know.

[NielsRogge]

Hi,

This was addressed in #27698. It's already fixed on main: https://huggingface.co/docs/transformers/main/en/model_doc/owlv2#transformers.Owlv2ForObjectDetection

[Itto1992]

Thank you for your quick response!
I will try the main branch😁