Start a new topic

Difference Between FR models - Windows SDK

We would like to know what is the difference between FR lite model and full model.


And also do you have the free test/evaluation SDK based on Windows PC?

1 Comment

The two models have different architectures. The lite model is optimized for running on resource constrained devices such as embedded systems. It has a lower memory footprint and also has faster inference time. The full model has better accuracy, but also requires greater memory and has a larger inference time. It is ideal for x86 systems with more computing power or systems with GPUs.

You can view the difference in accuracy here:
The full model is the "Model-TFV4", while the lite model is the "Model-lite".

The difference in inference speed and memory usage is summarized in the table on this page:

We have several different developer solutions which can run on different systems. 

For Windows, I would suggest either the Visionbox, which is a dockerized application exposing REST API endpoints, or the python SDK (which is now deprecated). 

You can learn more about the solutions at the following links.
Python SDK:

We are also currently in the process of adding Windows support for out main C++ SDK.

Login or Signup to post a comment