Google releases new TensorFlow Object Detection API

Google is launching its New TensorFlow Object Detection system. The Internet company has announced this in a blog contribution. Generally, the tensorflow uses the technology in applications like Google Photos and Google translators as well as for features such as search and intelligent answers. Now the new tensorflow object detection system uses the technology to detect objects in photos.

In general, the smattering of models consists of this sort of detection API include the intense origination supported coiled neuronic networks (Neural Network are the groups of algorithms that are constructed according to a human brain in order to recognize recurring patterns and then to label them. The recognized patterns are translated into mathematical vectors. All information from the real world, such as images, sound, text or timelines, is taken into account. Neural networks help to classify new information on the basis of similarities and group them into model groups over several levels. Labels help to identify these groups.


Examples: Spam, No spam, Satisfied customer, Unsatisfied customer, Purchased link, Unsold link) as well as efficient models which are predestined to work on little sylphlike gadgets. The single shot detector called MobileNets is honed to run in real-time on a phone. In detail, the program for the object recognition requires an exact specification and must know which of the individual settings (features) to be searched for in an image. In order to find only a certain type of object, a developer specifies the algorithm that extracts features from an image.

For example, for facial recognition, several features are defined which make up a human face. Usually the light-dark differences (hair cascades), which give the machine a template for the search. If the algorithm finds a certain number of matches with this template in an image, the image is classified as TRUE. Support vector machines (SVM) are usually used to classify the found features. OpenCV is a well-known library that summarizes many such object recognition procedures. For example, face detection can be implemented without great effort.

The most demanding area in image analysis is object classification. It requires an algorithm that is robust enough to recognize all features of an object, regardless of the background. And this is now achieved by Google with its New TensorFlow Object Detection system. This wonderful system can detect Face, Object, and even a Landmark. The new system is designed to be more flexible and easier to use, and more powerful than the first generation of machine learning software.

Google believed that the machine-learning community – anyone from academic researchers to engineers to hobbyists – can exchange ideas much more quickly by working on the code rather than just reading research reports. TensorFlow uses the computing power of CPUs as well as GPUs and runs on desktops, servers, and mobile devices. In some tutorials and flow charts, Google explains how the system works. TensorFlow is intended to simplify the implementation of software for machine learning.This, in turn, might speed up research on machine learning and ultimately will make the technology work better for everyone.

Other companies like Facebook or Microsoft are also working on the development of machine learning. Facebook has provided its tools for machine learning and artificial intelligence already in January of the open source community.


Tags : Artificial IntelligenceTensorFlow

The author FileEdge