Periodic Reporting for period 1 - 2DNEURALVISION (2D materials enabled Wide spectrum image sensors and optical NEURAL network based low power consumption computer VISION)
Reporting period: 2023-10-01 to 2025-03-31
The 2DNEURALVISION project aims to develop the enabling components for a novel low-power consumption, any weather, any light computer vision system. These components are a 2DM enhanced wide-spectrum image sensor and optical neural network with enabling 2DM passive and active elements. The 2DM in the wide-spectrum image sensor enables an increased signal to noise ratio for wide spectrum absorption at a low pixel power consumption. The processing of the image sensor data will be implemented through an ONN that offers a dramatic compute performance increase and power consumption reduction with respect to current vision processing systems thanks to 2DM enabled speed and bandwidth increase. Non-toxic absorber materials for the infrared spectrum and new waferscale back-end-of-line integration (BEOL) processes for 2DMs will be developed enabling compatibility of the 2DNEURALVISION technology with high volume markets.
The newly developed 2DM based technology components will be validated in an automotive use-case scenarios. The proof of-concept 2DM enabled low-power consumption vision system is projected to significantly outperform current solutions providing cost, performance, power, and form factor benefits to the end-users.
The newly developed 2DM based technology components will be validated in an automotive use-case scenarios. The proof of-concept 2DM enabled low-power consumption vision system is projected to significantly outperform current solutions providing cost, performance, power, and form factor benefits to the end-users.
The main achievements of the 2DNEURALVISION project include the development and successful testing of a functional QD camera in fog conditions, initial fabrication of an ONN chip, and progress in benchmarking its performance. Field tests highlighted the potential of SWIR cameras for enhanced pedestrian detection. Key advancements were also made in the design of graphene-integrated photonics platforms, packaging solutions for ONN deployment, and the implementation of an FPGA-based control system capable of real-time image processing.
The project's results have advanced the state of the art in 2D material integration, non-toxic quantum dot synthesis, and CMOS-compatible device fabrication. These achievements lay the groundwork for ultra-low-power, broad-spectrum image sensors with direct relevance for scientific, industrial, and automotive applications. The work has also resulted in new process flows, materials. and device architectures that are likely to yield high-impact publications and further innovation in the field. Continued progress is expected to deliver full system-level demonstration and further quantifiable advances by the next reporting period.