TensorFlow is the second-generation artificial intelligence learning system developed by Google based on DistBelief. Its name comes from its own operating principle. Tensor (tensor) means N-dimensional array, Flow (flow) means calculation based on data flow graph, TensorFlow is the calculation process of tensor flowing from one end of the flow graph to the other end. TensorFlow is a system that transmits complex data structures to artificial intelligence neural networks for analysis and processing.
TensorFlow can be used in a variety of machine learning and deep learning fields such as speech recognition or image recognition. The deep learning infrastructure DistBelief developed in 2011 has been improved in various aspects. It can be used in as small as a smartphone, as large as a few. Thousands of data center servers run on various devices. TensorFlow will be completely open source and anyone can use it.
The feature of TensorFlow is that it can support a variety of devices, from GPUs and CPUs to tablets and mobile phones. And it is very convenient to use, a few lines of code can start to run the model, which makes the introduction of neural networks very simple.
Among the many open source deep learning frameworks, tensorflow is also unique. Google’s name and appeal are of course one of the reasons, but TensorFlow also has many advantages. For example, the network structure code is very concise, the execution efficiency of the distributed deep learning algorithm, and so on.
As a "caiji", what attracts me most is that TensorFlow supports the python interface very well. After all, no data mining task can be done by simply defining the network. Numpy, Scipy, Pandas and other components owned by Python can greatly improve work efficiency.
TensorFlow programs are generally divided into two stages, the construction stage and the execution extreme. Generally, a graph is created in the construction phase to represent the neural network, and the training graph is repeatedly executed during the execution phase.
An artificial intelligence product launched by Google last year, the AIY artificial intelligence development kit, is a good tool for learning TensorFlow. The full name of AIY is Artificial Intelligence Yourself. As the name implies, it is a DIY function kit that uses AI. The goal of this project is to allow everyone to DIY their own AI artificial intelligence products and truly realize the civilianization of AI.
In order to facilitate users to learn and understand TensorFlow and artificial intelligence, Google also provides three TensorFlow-based neural network models for its AIY hardware-Vision Kit (it not only has a camera, but also supports neural network models), which can simultaneously identify Thousands of identical objects can be used to detect faces and expressions of people, or to identify targets such as people, cats and dogs. Users can also train the TensorFlow neural network model through the compilation tool provided by Google, allowing the AIY Vision Kit to recognize more things.
About AIY Vision Kit: AIY Vision Kit (vision kit) is a DIY artificial intelligence kit launched by Google that is worth $89.99. Google’s AIY Vision Kit allows you to build your own smart camera that can use machine learning to view and recognize objects. All of this is packed in a convenient small cardboard box, powered by Raspberry Pi and Intel movidius. It can run deep learning algorithms with low-consumption processing capabilities.
According to reports, the Google AIY series of development kits have arrived in Hong Kong for sale, and the prices are HK$706 and HK$392 respectively. Interested developers can pay attention to the Hong Kong online store of Google AIY series development kits-GravitylinkStore: https://store.gravitylink.com .
It can be connected to Raspberry Pi Zero WH via Raspberry Pi V2 camera. The board includes a vision processing unit chip that can efficiently run TensorFlow image processing. And this kit comes with a large number of sample models out of the box, which users can use offline.
In terms of machine learning frameworks, TensorFlow is truly unique in that it can build models in 5 or 10 lines of code. Then apply this model and expand to make a product. AIY can just help developers/beginners to practice, quickly familiarize themselves with TensorFlow, and then learn and develop.