Organizations beginning their computer vision journeys — including those leveraging Labelbox — often harness the power of YOLOv5 models to speed up the process of object detection to production. Recently, the Ultralytics team collaborated with Labelbox to release an exciting new integration that converts information in the JSON file format to the normalized bounding box formats that are required for YOLOv5 models.
Labelbox customers can now combine any of the YOLOv5 models with our training data platform to produce performant, production-ready object detection models faster than ever.
Less than a decade ago, machine learning was often inaccessible and required engineers to create models from scratch if they wanted to explore the opportunities offered by the emerging field. In 2014, Glenn Jocher founded Ultralytics with the purpose of democratizing ML. A few years later, the company launched YOLOv5, a family of open-sourced, compound-scaled object detection models trained on the COCO dataset.
YOLOv5 can be loaded from PyTorch and includes some of the most advanced off-the-shelf options for those starting their enterprise ML journey. The models are used today by a wide range of ML teams, including academic research, government agencies, and enterprises.
The new connector is open source and available on Github. If you want to learn more about the advanced capabilities of these models, check out the YOLOv5 repository. The Ultralytics team will also soon be launching the Ultralytics HUB, a platform that will provide robust, code-free tools to help the ML community import data, train models, and deploy them at scale. Sign up for an invite to beta the HUB on their website!