Google has released a new TensorFlow object detection API as part of the continued development of its TensorFlow framework. Willing to offer the best of simplicity and performance, it will provide developers and data scientists access to the same tech Google uses for its own system, such as Nest Cam, street number identification in Street View etc.
The models comprised in the TensorFlow object detection API contain heavy duty inception-based convolutional neural networks, as well as streamlined models that are created to operate on less sophisticated machines. Also, earlier this week, the tech giant announced its MobileNets family that is capable of handling tasks such as facial recognition, object detection, and landmark recognition.
The smartphones we use today don’t empower the computational resources of server-based setup and larger scale desktops. So, the developers are left with two choices; either they have to use the machine learning models to run the cloud, or they have to simply the models themselves.
Tech giants like Google, Apple, and even Facebook have been contributing into these mobile models. This year at WWDC, Apple unveiled CoreML, with an attempt to integrate machine learning into iOS apps. Also, at I/O Google released TensorFlow Lite, a lighter version of its streamlined machine learning framework. Besides, Facebook’s Caffe2Go framework can develop models to run on smartphones.
If you want the TensorFlow object detection API, download it from here.