Google has announced the TensorFlow Lite Model Maker. TensorFlow Lite is an open-source deep learning framework for on-device inference. The new tool is designed to adapt machine learning models to datasets with transfer learning.
“It wraps the complex machine learning concepts with an intuitive API, so that everyone can get started without any machine learning expertise. You can train a state-of-the-art image classification with only 4 lines of code,” Khanh LeViet, developer advocate at Google, wrote in a blog post.
The Model Maker supports models available on the TensorFlow hub such as the EfficientNet-Lite models. In addition, it supports image classification and text classification. The team plans to provide more support for computer vision and natural language processing use cases.
Google has also added new fields in the metadata to simplify on-device machine learning. The metadata fields fall under: machine-readable parameters and human-readable parameters.
Other updates the company has made to the open-source project include new pretrained models, a code generator to generate wrapper code, benchmark tools to measure model performance of models, and support for more platforms.
Going forward, the team plans to release up-to-date on-device models, publish new tutorials and examples, enhance Model Maker to support more tasks, expand metadata and codegen tools, and launch more platform integration.
“TensorFlow Lite is the official framework to run inference with TensorFlow models on edge devices. TensorFlow Lite is deployed on more than 4 billions edge devices worldwide, supporting Android, iOS, Linux-based IoT devices and microcontrollers,” LeViet wrote.
