Today, anyone with a phone has the means to capture video recordings of human rights abuses. But this proliferation of data can easily overwhelm those who want to use this visual evidence to hold bad actors accountable. For journalists, civil rights investigators, war crimes researchers, and other advocates, this often means that critical evidence goes untapped to protect human rights.
SurvAI uses cutting-edge motion and object detection technology to automatically detect and categorize violence in video. The platform dramatically speeds up and enhances investigative capability in cases of unlawful violence. The tool, while still in pre-release, is being used in investigations into the January 6 mob attack on the US election, state violence against Black Lives Matter protestors at Lafayette Square, while also being tested by international human rights and journalistic investigators.
Sean Backstrom is a Los Angeles-based deep learning engineer. After work in digital music production, Sean began dabbling in technology development, struck by the artistic and creative side of coding. After an intensive coding boot camp, Sean became involved in a project using natural language processing to document police violence on social media. Stuck on the question of how AI might better analyze video and not simply text for violence, he dove headlong into study of advanced computer vision models, and began building his own. According to Sean, he started out, “learning how to jam the biggest machine learning models into my fledgling GPU as I could.” Beyond success with his product, Sean hopes to open the eyes of AI engineers to how their talent can drive impact. “It’s often said in the machine learning community that AI can’t make a difference in areas like human rights and social justice, but that’s unequivocally wrong,” says Sean. “I want to inspire others to bring the most cutting edge technology into this sector of society.”