#250 · Primary category: Computer Vision

DINO

computer-vision deep-learning object-detection

[ICLR 2023] Official implementation of the paper "DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection"

Project last updated:07/31/24

GitHub Stars

2.8K

Forks

309

Contributors

8

License

Apache-2.0

Why we included this project

If your detection pipeline has to work outside the lab, on dashcams or factory floors or store shelves, DINO is worth a close look. It keeps the DETR-style end-to-end design, so you skip the hand-tuned anchors and post-processing that older detectors rely on, and instead gets its accuracy from denoising anchor boxes plus a contrastive approach to learning queries. The repo gives you training and evaluation scripts, checkpoints for several backbones, and a notebook to run inference and see the results, which makes it straightforward to move from the paper to your own data. And because this is the official code behind an ICLR 2023 paper, it doubles as a readable reference for how modern transformer detectors actually work.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category