Google has released its object detection and recognition system to the developer community, through its TensorFlow Object Detection API. It’s an open-source framework built into TensorFlow, which Google says ensure object detection systems are easy to create and train.
In a blog post announcing the release, Google talks about how the framework has already been used in-house:
“Last October, our in-house object detection system achieved new state-of-the-art results, and placed first in the COCO detection challenge. Since then, this system has generated results for a number of research publications and has been put to work in Google products such as NestCam, the similar items and style ideas feature in Image Search and street number and name detection in Street View.”
It’s noted the Single Shot Multibox Detector with MobileNets — a core component of the system — is lightweight enough for mobile device use, plus there are Faster RCNN models for more powerful devices. Google says it welcomes contributions to the codebase, and will also add further updates in the future.
Visit Google’s website here to download the code.