Skip to content

lush-tech-warriors/lens-edge

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Lush Digital logo

Lens Edge Library

This is a small wrapper library that can be used to run inference on a TFLite model in Python with the option of using a Coral EdgeTPU for faster inference times.

The library uses the TFLite runtime package as a full TensorFlow installation includes a lot of functionality that is not required to run inference. Because of this, there is a requirement of Python version 3.5 or 3.7 on an armhf architecture.

Setup

If a Coral EdgeTPU device is going to be utilised, firstly follow the installation process for the EdgeTPU runtime, which can be found on the getting started pages.

Once this is complete, install this package using the following

pip install https://github.com/LUSHDigital/lens-edge/archive/master.zip

Usage

This is a super easy to use wrapper, so, the basic usage looks rather small. There are some more examples in the examples directory that can be looked at. But the basic use to run inference on a CPU is

import lens_edge

# Run inference on CPU
model = lens_edge.infer('TFLITE_MODEL_PATH', 'LABELS_PATH')
results = model.run('IMAGE_PATH')

print(results)

to run inference on a Coral EdgeTPU accelerator, limiting to one return match with a minimum threshold of 50% use

import lens_edge

# Run inference using Coral EdgeTPU adjusting count and threshold
model = lens_edge.infer('EDGETPU_TFLITE_MODEL_PATH', 'LABELS_PATH', 'libedgetpu.so.1')
results = model.run('IMAGE_PATH', count=1, threshold=0.5)

print(results)

Credits

License

The MIT License (MIT). Please see License File for more information.

About

Lush Lens EdgeTPU

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages