Face Forensics  Image Recognition Suite
Advanced Face Recognition           Tattoo Recognition
Partial Face Recognition                 Scene Recognition

     ChildScene:    Automatic processing of scenes of child abuse

The ChildScene system is the result of the Face Forensics team’s previous experience in developing unique AI-based software to automate the processing of seizures containing images of child abuse. It substantially reduces the time to process a seizure, and the trauma experienced by the investigators who have to view the images.

•ChildScene compares images of abuse on seized PCs or websites against a master database of images and videos from previous seizures, distinguishing between known and unseen images (even if heavily cropped or reformatted), enabling investigators to focus on the unseen images which can often comprise less than 5% of a seizure

•It can lead to the possible recovery of the children, and the identification of offenders.

•It links images of scenes of abuse on seized PCs with victims, offenders, tattoos, and specific objects in the images, helping identify victims and offenders through face, partial face, tattoo, and object recognition

ChildScene has been designed to connect easily to existing law enforcement databases containing known images and videos of abuse. It accesses these in read-only mode. The system holds no images itself, just the encode arrays. These cannot be reversed back to the images.

ChildScene Database Relationships
An Image (on the Right) Being Matched Against a Database of Full and Cropped Images in Different Formats.  Matches are on the Left

The primary use of ChildScene is to automatically separate new images of abuse on seized media PCs, flash drives, websites, and other sources from known ones. To do this the media is copied and then connected as an external drive to a fast multi-core processor array running ChildScene. The system will read each image or auto-selected video frame, encode it using its unique algorithms, and compare each against a master database containing all the encode arrays of images that have already been seen in previous seizures. Unlike existing image matching approaches, ChildScene will recognize images that have been resaved, reformatted, or that have been lightly or heavily cropped. Previously unseen images (which are likely to be a small subset) may well be of children that the suspect has abused. They are copied to a separate folder for officers to investigate.

All previously unseen images above a selected severity level can optionally be processed using face, partial face, object, and tattoo recognition to match against databases of victims and offenders, enabling victims to be recognized and offenders identified. Within ChildScene the core entity is a scene, i.e. an image of abuse.

Key features:

•The matching speed of an encoded scene depends on the hardware configuration but can be in the millions of images per second. ChildScene can be easily networked, allowing users to connect to different databases on a LAN or VPN under appropriate controls The master database can hold a virtually unlimited number of records.

•ChildScene has been designed to work in conjunction with existing third-party software

•ChildScene is built on experience gained by the team in developing the ChildBase system for the UK National Crime Squad (now the NCA/CEOP). This was launched at the House of Lords.

System Requirements: ChildScene is client/server software available as a stand-alone/networked application and as a .NET SDK or web service. It runs under Windows 10 64-bit, and will access Oracle, SQL Server, and DB2 databases. The user interface can be displayed in any language. ChildScene will recognize scenes and faces in still images, video files, and websites.

ChildScene is available to law enforcement agencies for evaluation at no charge.

For more details law enforcement agencies should contact info@faceforensics.com