Most Popular Deep Learning Frameworks (2022)

Last updated on January 1, 2022 Guest Contributor Loading... Post a comment

It is difficult to determine which is the best among deep learning frameworks because each has its own strengths in different scenarios. However, it is quite possible to know which deep learning frameworks are trendy and used the most by big companies, startups, and independent software developers. The things that are used a lot will bring many benefits and convenience to you, such as:

  • There will be more jobs.
  • There will be more documentation, more tutorials from blogs, more questions and answers.
  • There will be many free libraries related to them that you can use right in your projects.
  • These open-source libraries will be contributed more, maintained better, and updated regularly.

This article will walk you through a list of the most popular deep learning frameworks in 2022, based on the number of stars on GitHub and the number of new questions about them on StackOverflow.


Tensorflow has a number of stars on GitHub and the number of related questions on Stack Overflow outperforms other deep learning frameworks. Although Tensorflow 1.x is very complicated and troublesome to implement, Tensorflow 2.x is very user-friendly and eliminates the clutter. Now, you can build and train machine learning models easily using its intuitive high-level APIs.

With Tensorflow, you can train or deploy your models for production easily thanks to Tensorflow.js (for Node.js and browsers), Tensorflow Lite (for Android and iOS), etc.


Keras offers consistent and simple APIs as well as minimizes the number of actions required for common use cases.

Keras has a very close relationship with Tensorflow. In fact, it is built on top of Tensorflow 2.0. You can export Keras models to JavaScript to run directly in the browser, to TF Lite to run on iOS, Android, and embedded devices.


PyTorch is primarily developed by Facebook’s AI Research lab. It is based on the Torch library and focuses on computer vision and natural language processing. Many tech giants use PyTorch including Tesla and Uber.

The framework is built to be deeply integrated into Python and you are free to reuse your favorite Python packages such as NumPy, SciPy, etc, to extend PyTorch when needed.

The PyTorch Mobile runtime provides an end-to-end workflow that simplifies the research to production environment for mobile devices. You can easily bring your trained models to Android and iOS apps.

Apache MXNet

  • GitHub stars: 20k+
  • Maintained by: Apache Software Foundation
  • Written in: C++, Python, Cuda, CMake, Shell
  • Links: GitHub repo | Official website

MXNet is designed for both efficiency and flexibility. It allows you to mix symbolic and imperative programming to maximize efficiency and productivity. The framework is lightweight, memory-efficient, and portable to smart devices through native cross-compilation support on ARM, and through ecosystem projects such as TVM, TensorRT, OpenVINO.

Every one who are familiar with NumPy can easily learn and use MXNet.

Old Deep Learning Frameworks

You may be wondering that there are deep learning frameworks that are very popular but do not appear in the list above. The answer is because they have completed their lifecycle or because they are no longer active development, fix bugs and add new features. Some typical examples:

  • Microsoft Cognitive Toolkit (CNTK)
  • Caffe, Caffe2
  • Torch


We have covered the most beloved deep learning frameworks in the year 2022. Spend a few hours playing around with them and select the one you’ll be traveling with. Once you master this, it’s easier to master the next because there are some similarities between them.

If you would like to explore more things about Python and machine learning, check out our Python category page for the latest tutorials and examples.

Notify of
Inline Feedbacks
View all comments

Related Articles