While Microsoft has expanded the Bing image search toolset to allow searching for specific items within a larger image, Google has followed suit by releasing a new TensorFlow object detection API to make it easier for developers and researchers to identify objects within images.

The detection API comprises of convolutional neural networks, with streamlined models designed to run on less-powered devices, and MobileNets single shot detector optimized to run in real-time on smartphone.

With the MobileNets family of lightweight computer vision models designed to handle tasks like object detection, facial recognition and landmark recognition.

Albeit, smartphones possess less computational resources than desktop and server-based setups, which leaves developers with only two options - machine learning models based in the cloud, which adds latency and requires internet connection, and the alternative approach which is simplifying the models, is a trade-off in the interest of more ubiquitous deployment.

With Google's offerings in public cloud services to give it differentiated and advantage positioning, as obviously, Google isn't new to delivering computer vision at scale in relation to its Cloud Vision API.

Google has made it extra easy to play with and implement, with the addition of Tensorflow detection model definitions now available on Github.

TensorFlow object detection API for identifying objects within images



While Microsoft has expanded the Bing image search toolset to allow searching for specific items within a larger image, Google has followed suit by releasing a new TensorFlow object detection API to make it easier for developers and researchers to identify objects within images.

The detection API comprises of convolutional neural networks, with streamlined models designed to run on less-powered devices, and MobileNets single shot detector optimized to run in real-time on smartphone.

With the MobileNets family of lightweight computer vision models designed to handle tasks like object detection, facial recognition and landmark recognition.

Albeit, smartphones possess less computational resources than desktop and server-based setups, which leaves developers with only two options - machine learning models based in the cloud, which adds latency and requires internet connection, and the alternative approach which is simplifying the models, is a trade-off in the interest of more ubiquitous deployment.

With Google's offerings in public cloud services to give it differentiated and advantage positioning, as obviously, Google isn't new to delivering computer vision at scale in relation to its Cloud Vision API.

Google has made it extra easy to play with and implement, with the addition of Tensorflow detection model definitions now available on Github.