Hey everyone!

I'm working on an object-detection project and fine-tuning Meta's DETR model on a custom dataset of around 500 images.

The main challenge I'm facing right now is annotation. I need to label multiple object classes with bounding boxes, and manually annotating every image is becoming a time-consuming process.

I want to speed things up without sacrificing accuracy, since poor annotations could affect the model's performance.

I've been looking into tools and models such as CVAT, Roboflow, Grounding DINO, and OWLv2 for AI-assisted annotation.

My current idea is to use a pretrained detection model to generate initial bounding boxes, manually review and correct them, and then convert the annotations into the format required for fine-tuning DETR.

However, I'm unsure which workflow would be most reliable for my specific use case.

I'd appreciate advice on the following:

What is the best annotation tool for labeling 500 images accurately and efficiently?

Would Grounding DINO or OWLv2 be suitable…