Google has announced the release of the TensorFlowObject Detection API, which is an open-source framework built on top of the TensorFlow open-source software library for machine learningthat makes it easy to construct, train and deploy object detection models, according to the company.
Last October, Google noted that its in-house object detection system achieved new, state-of-the-art results, and placed first in the COCO detection challenge. Since then, the system has generated results for a number of research publications, and has been put into work in Google products such as NestCam. Thus, Google chose to make the codebase available to the broader research community. The goal of the company was to support cutting edge models while allowing for rapid exploration and research.
The first release contains a selection of training detection models, which Google hopes that the open-source community will find useful for its computer vision needs. Contributions to the codebase, notes the company, are also welcome. Google also suggests users stay tuned for further updates to the framework, and that interested users can now download the code and try detecting objects in some of their own images using the Jupyter notebook, or training their own pet detector on Cloud ML engine.
View more information.
Learn more: search the Vision Systems Design Buyer's Guide for companies, new products, press releases, and videos