Stage 1
15:00 - 16:00
Live Translation
Accelerating Human Rights Investigations with Computer Vision

Short thesis

This session will present VFRAME, a computer vision toolkit being developed in collaboration with the Syrian Archive initiative to accelerate information gathering capabilities for investigative journalism. VFRAME (Visual Forensics and Advanced Metadata Extraction) enables researchers and journalists to detect objects of interest, classify scenes, extract text, filter graphic content, and create training databases for building custom recognition models.


The Syrian Archive is a collective of human rights activists dedicated to documenting human rights violations and war crimes committed by all sides in the Syrian conflict. Their network of researchers and journalists are pioneering new methods for accountability and advocacy, which rely on collecting, verifying, and annotating thousands of videos posted online that contain evidence of human rights violations. But manually reviewing these videos is expensive, does not scale, and cause vicarious trauma. Currently there are over 1 million videos waiting to be reviewed.

The VFRAME computer vision toolkit is being developed in collaboration with the Syrian Archive initiative to improve their information gather capabilities. VFRAME can analyze large-scale video datasets to locate illegal munitions (e.g. cluster bombs, incendiary weapons, barrel bombs); filter graphic content; and extract metadata about scenery, logos, and text. This metadata is then made searchable through both text and visual queries allowing the Syrian Archive to more quickly locate and verify videos that contain evidence of war crimes and human rights violations.

This session will discuss and showcase how VFRAME and the Syrian Archive initiative are working together to innovate new tools and improve open source information gathering technology to strengthen justice and advocacy efforts.