CORTEX2 innovators: XR-CARE's 1st progress update

Q: What is XR-CARE in one sentence?

A: XR-CARE is a modular, multimodal anonymisation framework designed to ensure privacy in XR teleconferencing by detecting and obfuscating sensitive information in video, audio, and text data.

Q: What problem are you solving? What makes your solution unique?

A: With XR-CARE, Logimade is addressing the growing privacy risks associated with recording and sharing XR teleconferencing sessions, where sensitive personal information—such as faces, voices, and on-screen text—can be unintentionally exposed. Current anonymisation tools are either limited to single data modalities or are too slow and complex for practical use in large-scale deployments.

What makes our solution unique is its modular, multimodal, and multi-stage architecture, which allows for configurable, high-recall anonymisation of video, audio, and text streams. It operates efficiently on consumer-grade hardware, supports real-time anonymisation, and adapts to different teleconferencing contexts (e.g., XR, desktop, mobile). Additionally, our system emphasises usability and transparency, allowing users to customise parameters, track processing history, and optimise anonymisation based on their needs—something no off-the-shelf solution currently offers.

Q: What are XR-CARE’s main objectives?

A: 

1. Develop a multimodal anonymisation framework capable of processing video, audio, and text from XR teleconferencing sessions while preserving contextual usability.
2. Ensure high privacy protection through a multi-stage detection strategy that minimises false negatives and adapts to varied teleconference scenarios.
3. Enable near real-time anonymisation using consumer-grade hardware, making the solution practical and scalable for widespread adoption.
4. Provide a user-friendly web platform where users can manage projects, customize anonymization parameters, and track processing history.
5. Integrate XR-CARE anonymisation platform with CORTEX2 framework.

CORTEX2 support programme progress

Q: What were the main activities implemented and milestones achieved during Sprint 1 of the CORTEX2 Support Programme?

A:

1. Development infrastructure was successfully established, including GPU-enabled environments, version control, and benchmarking tools to support reproducible research and AI-based processing.
2. Multi-scenario datasets were collected and annotated, capturing diverse XR teleconference conditions across video, audio, and text modalities, including challenging cases such as occlusions and varied lighting.
3. Comprehensive benchmarking of face and body detection algorithms was performed, evaluating models like YOLOv10 and MediaPipe for accuracy, speed, and robustness in realistic teleconferencing scenarios.
4. Initial evaluations of text and voice detection and obfuscation techniques were completed, with comparative tests of models such as FAST for text and Silero VAD for speech, along with analysis of obfuscation methods including Gaussian blur, pitch shifting, and spectral modification.
5. A modular software architecture was defined, enabling configurable, multi-stage anonymisation pipelines with support for adaptive processing and multimodal data integration.

Q: What have you achieved so far?

A: So far, XR-CARE has successfully progressed from the solution design and component evaluation (Sprint 1) to delivering an integrated, tested, and multi-modal anonymisation solution ready for deployment (Sprint 2).
During Sprint 2, three major milestones were achieved:

1. Integration with the CORTEX2 Platform: The team developed a production-ready RESTful API and a user-friendly web interface, enabling seamless access to and control of the anonymisation pipeline. This integration lays the groundwork for scalable, real-world application of the XR-CARE system within the CORTEX2 ecosystem.
2. Extension of Multi-Modal Anonymisation: The system was expanded to support visual anonymisation of health-related IoT sensor data embedded in video recordings, reinforcing the framework’s robustness in healthcare contexts and enhancing its capacity to process diverse XR data streams.
3. Comprehensive Testing and Debugging: Extensive testing was carried out using real-world XR teleconference datasets to validate performance, optimise speed and accuracy, and ensure compliance with GDPR. The resulting framework demonstrated high recall, low false positive rates, and efficient processing on consumer-grade hardware.

These developments have transformed XR-CARE into a practical, multimodal anonymisation solution capable of supporting privacy-preserving XR teleconferencing. The project has increased the team’s technological readiness and positioned XR-CARE for final validation and deployment in real-world CORTEX2 use cases.

Q: How is participating in CORTEX2 supporting XR-CARE?

A: The most valuable aspect of CORTEX2 support has been the technical mentorship provided by Alireza Javanmardi, whose guidance has been critical in refining our multi-stage anonymisation strategy. His input helped us make key architectural decisions, while the teleconference recordings from Open Rainbow he shared enabled more realistic and rigorous validation of our system.

Additionally, the encouragement to submit a full article and poster to EuroXR 2025, along with financial support for conference participation, has provided an excellent opportunity for dissemination, visibility, and networking. This not only helps to promote our work but also opens doors for potential collaborations and economic exploitation of the XR-CARE platform.

Q: What are your next steps within the CORTEX2 Programme?

A: The next steps of the project XR-CARE focus on bringing the framework to ready for public release through real-world testing, refinement, and dissemination:

1. Real-World Validation: We will conduct validation trials in real healthcare teleconferencing scenarios to assess the framework’s effectiveness. This phase will include the collection of performance metrics and user feedback to evaluate the system’s robustness, usability, and compliance with privacy standards.
2. Final Optimisation: Based on insights gathered during validation, we will implement targeted improvements to the anonymisation pipeline. Particular attention will be given to optimising detection performance, reducing processing time, and incorporating feedback from healthcare professionals and other end users.
3. Promotion and Dissemination: We will conduct a final review to ensure all project objectives have been fulfilled. In parallel, we will prepare promotional materials, including presentations, documentation, and online content, to support the visibility and potential adoption of XR-CARE beyond the CORTEX2 program.

Learn more about XR-CARE and stay updated on its progress!

Want to explore more XR innovation? Browse all our supported projects on the CORTEX2 website: 

Open Call 1 winners  –  Open Call 2 winners