Digital Archiving & Documentation Operations
Alongside our published investigations, we run a partner-supported digital archiving operation documenting events and potential human rights violations in North Africa and East Asia. Between October 2023 and June 2026, we monitored 246 channels and collected 903 satellite images to process a massive volume of digital evidence.
Of the extracted multimedia: 16,992 JPG images and 9,332 MP4 videos.
For inquiries or to request access to this archive, contact Archive@paxmemoria.org.
Strategic Purpose & Capacity Building
We do not store geographic information systems (GIS) and multimedia data permanently. Our primary objective is capacity building: working through partners, we act as a bridge, equipping local civil society organizations, human rights groups, and UN agencies with the exact data, analytical tools, data-management practices, and workflows they need to document violations independently.
Narratives Beyond Criminal Justice
We use this archived data to understand "information flow." Our vision extends beyond using evidence solely for criminal justice and legal proceedings — we explore its potential to analyze, understand, and dismantle narratives, helping preserve collective memory and document realities in ways that transcend strict legal frameworks.
AI Triage & Classification
We integrate AI models into our workflows to rapidly classify massive datasets. These models analyze images and video to automatically flag five categories relevant to potential violations:
Destruction
Structural damage to civilian infrastructure and residential areas.
Casualties
Presence of victims, wounded individuals, or medical evacuations.
Military Presence
Armed individuals, military personnel, tanks, APCs, and military bulldozers.
Hazards
Active fires, smoke plumes, strike impacts, weapons, and unexploded ordnance (UXO).
Environmental Damage
Deliberate destruction of agricultural lands and changes in vegetation cover.
GIS Integration & Strict Manual Verification
While AI accelerates data processing, the final comparison, verification, and issuance of results are conducted entirely manually. Our team cross-references collected visual media with GIS satellite imagery to confirm geolocations, verify chronologies, and build accurate conclusions free from machine bias.
As a practical application of our spatial-analysis capabilities, we used this methodology for a damage assessment of the Syria wildfire. By combining ArcGIS imagery services, Open Data, and Planet Labs satellite imagery with SAM3 and DINOv3 deep-learning models, we rapidly converted a half-dozen point samples into a preliminary damage layer — which was then passed to our team for comprehensive manual review before any final results were issued.