VALID | My Pictures Matter

The VALID (Veracity, Agency, Longevity, Integrity and Dignity) project is developing frameworks that improve data quality, ethical accountability and build public trust in machine learning projects that seek to provide law enforcement and community safety outcomes.

The My Pictures Matter Campaign is a world-first and global crowdsourcing initiative launched by AiLECS Lab, to create an ethically sourced dataset of 100,000 ‘safe’ childhood photos for machine learning research to counter child exploitation. Instead of seeking proxy consent from parents or guardians, it enables people who are now adults (18 and above) to share their childhood photos with their own consent (and that of each person depicted) and withdraw it if they no longer wish for it to be used.

This dataset has been produced under the VALID project framework. These are images of children in benign and every-day contexts that help train AI models intended to detect child exploitation material. By using these images, AI models can learn what 'safe' images of children look like and, when combined with other datasets, can learn to tell the difference between ‘safe’ images of children from exploitative one.

This is critical to help develop tools that can quickly and accurately classify high volumes of digital image collections – including ‘new’ material that would not be detected through traditional methods like hash matching, which only identifies previously known images.

This project has been approved by the Monash University Human Research Ethics Committee (Project #31436)

Previous
Previous

Airwave | Voice Profiling

Next
Next

The Survivor's Perspective