System for automatic blurring of video streams from video surveillance cameras
5 486 M€
Number of employees
Setting up an anonymized video database following GDPR data protection requirements for AI experimentation
RATP is an expert in sustainable mobility and one of the 5 largest urban transport operators in the world. RATP operates 14 metro lines, 2 RER lines, 7 tramway lines and 350 bus lines in the Paris region. Carrying more than 12 million passengers a day, it is the largest multimodal network in the world. The RATP group conceives, designs and carries out infrastructure development projects, operates and maintains networks for any mode of transport (metro, regional train, tramway, bus), and develops innovative mobility assistance services (passenger information, remote ticketing, pricing, customer marketing).
- RATP operates 50,000 video surveillance cameras throughout the Île de France network.
- In this context, new technologies such as Edge Computing and Deep Learning represent new tools to exploit this capital and extract information from video streams.
- Video streams include personal biometric data and are subject to regulatory and legal constraints, including the General Data Protection Regulation (GDPR).
- To create a dataset to train AI models for smart video analysis and ensure their effectiveness in dense and complex environments such as public transport environments.
- To comply with the regulatory requirements of the GDPR on the use of videos containing personal data. Indeed, the CNIL prohibits the storage of videos containing biometric data for more than 72 hours, which makes the data impossible to use in the context of experimentation.
“Our main challenge is to optimize the use of our video data while ensuring the protection of our agents’ and travelers’ identity. The system developed by Deepomatic for RATP allows, through automated anonymization, to create video sequences without personal data that can be used in smart video experiments.”
SOLUTION : A real-time and irreversible anonymization tool
Deepomatic has developed with RATP a real-time anonymization tool. In order to remove all traces of personal data, video streams from video surveillance cameras are processed by detection algorithms that locate biometric data subject to GDPR, such as head, hair, and face. The identified elements are then anonymized by an irreversible blurring tool.
- Acquisition of a video database that can be used to train artificial intelligence algorithms.
- Reduction of the workload during the CNIL’s Privacy Impact Assessment (PIA), and guarantee provisions for the protection of personal data.
What are the next steps?
- Evaluate the Deep Learning algorithms on safety and operational use cases.
- Evaluate the feasibility of embedding video recognition algorithms at the sensor level for localized computing (Edge Computing)
“This project represents a first step in an innovation roadmap in areas such as safety where AI algorithms could be integrated into the business process by reducing the burden on operators.”