2021 Google Developers Conference丨Building an efficient machine learning ecosystem

Google is committed to building an advanced machine learning ecosystem,

Bring an efficient development experience,

And help developers to apply machine learning to multiple fields.

From scientific research, business transformation to public welfare,

Solve various problems in real life.

Are you ready?

Let’s get to know the 2021 Google Developers Conference

The latest tools and progress of TensorFlow

to go directly to the official website to explore more and develop dry goods!

This year TensorFlow brought a big surprise to developers-TensorFlow Decision Forests (TF-DF). The TF-DF model can achieve efficient classification, regression and task sequencing. When developers need to process tabular data for simplicity, interpretability and speed, TF-DF will be the best choice. TF-DF provides a large number of tools and models for developers to use. As a member of the TensorFlow rich ecosystem, it can be integrated with various TensorFlow tools, libraries, and TFX platforms, allowing developers to easily combine and use.

enter the Google developer online course, search for "TensorFlow" to master more development skills!

TensorFlow.js is a front-end machine learning library written in JS language. Because front-end machine learning can achieve low-latency running speed, protect user privacy, and lower deployment and maintenance costs, TensorFlow.js has also become the first choice of many developers. This year, the number of weekly downloads has reached 3 times the previous, and the total downloads The volume has reached 3.2 million times.

TensorFlow.js has flexible cross-platform features. Developers can run models at any time in a browser or any environment that supports JavaScript, giving full play to the advantages of wide web coverage and large scale.

TensorFlow.js has a rich model library and API, which is convenient for developers to quickly build applications. A new pose-detection API was launched this year, which can realize single-person detection and multi-person detection; two MediaPipe BlazePose models were released, which can be used in 2D and 3D scenes. In terms of natural language processing, BERT-based dialogue intention detection model and question answering model have also been introduced.

It is worth mentioning that TensorFlow.js already supports the TFLite model, which has faster execution speed, smaller model and better performance. The performance testing tool also adds the option of custom model, which can test and display many relevant indicators such as the execution speed of the model in real time.

58.com used car team, InSpace, IncludeHealth, Project Shuwa, etc. applied TensorFlow

Recommendation systems are an important application in the field of machine learning. From recommending movies and restaurants to recommending videos or news articles, recommendation systems are common in daily life. The recommendation system is a complex machine learning system that is divided into three stages: recall, rough ranking, and fine ranking, reducing the number of recommended content from millions to dozens of valuable recommended content.

A reliable and powerful recommendation system can greatly increase user activity, and TensorFlow Recommenders can become the best assistant for recommendation system developers.

enter the Google developer online course, search for "TensorFlow" to master more development skills!

Google has developed a series of principles to guide the understanding of responsible AI. AI should be beneficial to society, fair, safe, and protect privacy, and be responsible to users.

Google has launched a responsible AI toolkit to help developers develop AI in a responsible way, make progress, and build an AI system that benefits everyone.

The development of TensorFlow is inseparable from the support and contribution of the developer community. At present, the number of TFUGs has grown to more than 70, and more than 170 GDEs and 12 special interest groups (Sigs) have been born.

In terms of learning resources, in addition to the localized online courses provided by NetEase, developers can also master TensorFlow-related skills in the form of videos, articles, and Codelabs through Google developer online courses.

If you want to study further, welcome to take the TensorFlow Developer Certification Exam, get the TensorFlow Developer Certificate, and show your strengths. There are currently 3,000 certified developers, and we hope that more developers will join and the community will continue to grow.

Click to learn about TensorFlow Developer Certificate

Click to learn about the Chinese University MOOC "TensorFlow Introductory Practical Course"

In addition to the above wonderful updates, what latest developments has the TensorFlow team brought?

Log in to the official website to explore now!