Skip to content

A python interface to interact with the great Machinebox/Facebox

License

Notifications You must be signed in to change notification settings

jhormigo/facebox_python

 
 

Repository files navigation

facebox_python

License: MIT Build Status

A python interface to interact with the great Machinebox/Facebox. At the moment I have implemented only two endpoints:

1. facebox/teach
2. facebox/check

The reason I have written this interface is because I wanted to teach the facebox multiple people together, and because I wanted to get the result of recognition task facebox/check as an image with boxes around the detected faces.

Usage:

  1. Signup for a free account on Machinebox/Facebox.
  2. Run the docker-image using

docker run -p 8080:8080 -e "MB_KEY=$MB_KEY" machinebox/facebox

Don't forget to add the MB_KEY to your path or ~/.bashrc file on Linux. Check documentation on how to do that.

  1. from the commandline call either teach.py or check.py

Teach endpoint

To teach the facebox a person/face you have to make a folder with the name of the person. The dataset structure should look like this:

Dataset structure
person_1/
    img1.jpg
    img2.jpg
    ........
person_2/
    image1.png
    image2.png
    ..........
person_3/
    image1.jpeg
    image2.jpeg
    ..........    
..........

Allowed extensions for training/teaching are: .jpg, .png and .jpeg

Remember with a free license of Machinebox/Facebox you are allowed to teach only 20 faces.

Then just call teach.py from the command line At the end of the teaching process the file failed_log.txt will be written down. The file keep tracks of all failed teaching images/faces. So you can then check why an image has failed to be recognized or taught!

Check endpoint

You can the file check.py --file [filename_path] --save [yes or No]

--file argument:

you need to specify the iamge/path to the image you want to check

--save argument:

this is used to save the returned image from the facebox. If you give it yes it will save a copy the returned image with recognition boxes on the disk with the name machinebox-ORIGINAL_FILENAME.jpg. Or give it no and it will only display the returned image without saving it.

For example if you want to test the image trump1.jpg then you have to execute this command py check.py --image trump1.jpg --save no I don't want to save the image, I just want to see it

Note: py command is shortcut for my python3.6 on Ubuntu 18.04

License

MIT License

About

A python interface to interact with the great Machinebox/Facebox

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%