YOLO: Open source real-time image recognition

For those of you who are looking to play around with image recognition in your UAS projects, there’s an open source real-time image recognition system for that. I am not sure yet how well this would work at longer distances with smaller images when capturing footage from a flying platform but could be interesting. I wonder if SAR and related ops could perhaps benefit from such an open source project. Gene?

From Joseph Redmon’s (one of the creators) website: “You Only Look Once (YOLO) is a state-of-the-art, real-time object detection system. On a Titan X it processes images at 40-90 FPS and has a mAP on VOC 2007 of 78.6% and a mAP of 48.1% on COCO test-dev.”

Joseph’s recent TED talk on the system is well worth a watch:

And another one where it appears as though they are speed testing the recognition by letting it run on top of a James Bond action scene:

You can find the project’s paper, YOLO9000: Better, Faster, Stronger – by Joseph Redmon & Ali Farhadi on the Cornell University Library site: arxiv.org/abs/1612.08242

Or download it direct arxiv.org/pdf/1612.08242.pdf.

Joseph’s website has instructions for getting the system running: pjreddie.com/darknet/yolo/