Add Object Detection notebook (Images/Videos)

[hp6318]

Object Detection on Images/Videos with DINOv3

This notebook demonstrates using DINOv3 for object detection using a frozen DINOv3 Backbone and
a lightweight DETR style decoder trained on COCO dataset (as mentioned in DINOv3 paper). We will use
the pre-trained weights provided by the authors.

Given:

• RGB video frame(s)

We will extract bounding boxes with their scores and labels from each frame.

[lartpang]

Thank you for your contribution. It's truly very valuable.
It would be even better if the official team could provide inference and evaluation scripts for all tasks covered in the paper, such as segmentation, classification, detection, depth estimation, and so on.

[foxyglue]

This is great work, thank you so much for sharing this object detection notebook!

I'm looking to fine-tune this on a custom dataset and had a quick question. My dataset is in YOLO format. Would you recommend converting it to COCO format to work with this model, or is there a more direct approach you might suggest?

Thanks again for the excellent contribution!