Skip to content

Tools for Ethical Implementations of Machine Leanring

Notifications You must be signed in to change notification settings

iman-saleh/ethical-ai

Repository files navigation

This repository contains artifacts that support work on Ethical AI

License Plate Blurring Script

The blur-license-plate-tutorial.ipynb demonstrates capabilities to detect and blur license plates in images.

The scripts comes with sample images for testing. The images are loaded from 'license-plate' directory and the processed (blurred) images are saved under 'license-plate-processed' directory.

The script is tested on Ubuntu 16.04 and Python 2.7

Contact

Iman Saleh [email protected]

License

Affero GPLv3 http://www.gnu.org/licenses/agpl-3.0.html

Face Blurring Script

The blur-face-tutorial.ipynb demonstrates capabilities to detect and blur faces in images.

The scripts comes with sample images for testing. The images are loaded from 'pedestrian; directory and the processed (blurred) images are saved under 'pedestrian-processed' directory.

The script is tested on Ubuntu 16.04 and Python 2.7 and requires Tensorflow, Numpy, OpenCV and Pillow.

Contact

Cory Ilo [email protected]

License

Affero GPLv3 http://www.gnu.org/licenses/agpl-3.0.html

Bias Utility Script

The Bias_Detection.ipynb demonstrates capabilities to measure bias in the model.

The script downloads images and models hosted in [AWS S3] (https://s3-us-west-1.amazonaws.com/strata-bias-data2/inputdata.tar.gz). The script runs detection on images and models in inputdata directory. The script then calculates mean difference and unexplainable mean difference between pedestrian detection accuracy at day time vs night time images. The detected images are stored in outputdata directory.

The script is tested on Centos 7.6 and Python 3.6 and requires Tensorflow, Numpy, OpenCV and Pillow.

Contact

Cindy Tseng [email protected]

License

Affero GPLv3 http://www.gnu.org/licenses/agpl-3.0.html

About

Tools for Ethical Implementations of Machine Leanring

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published