2DNeuralVision is developing a low‑power, wide‑spectrum 2‑dimensional image sensor combined with an integrated optical neural network built from graphene and transition‑metal dichalcogenide materials. The system captures visible to short‑wave infrared light and performs early‑stage image processing optically, enabling robust vision in low‑light and adverse weather while reducing energy consumption for automotive, AR/VR, robotics, and mobile device applications.
Funding
Funding not disclosed
Founders
Product
Problem
Current computer vision systems for automotive, AR/VR, robotics, and mobile devices often rely on power‑hungry sensors and processing units that perform poorly under adverse weather or low‑light conditions, limiting reliability and energy efficiency.
Solution
2DNeuralVision is developing a low‑power, wide‑spectrum 2‑dimensional image sensor combined with an optical neural network (ONN) built from graphene and transition‑metal dichalcogenide (TMDC) materials. The sensor captures visible to short‑wave infrared light, enabling robust perception in low‑light and harsh weather. Integrated directly with the ONN, the system performs early‑stage image processing optically, reducing the computational load on electronic processors and cutting overall power consumption. The project delivers a compact, environmentally friendly component suite that can be integrated into automotive cameras, AR/VR headsets, service robots, and mobile devices, providing enhanced vision capabilities while supporting greener digital supply chains.
Target Audience
Primary customers are OEMs and system integrators in the automotive, augmented/virtual reality, service robotics, and mobile device markets that require energy‑efficient, high‑performance vision sensors for operation in low‑light or adverse weather environments.
Features
- Wide‑spectrum 2‑D image sensor sensitive from visible to short‑wave infrared, using graphene and TMDC layers for high quantum efficiency.
- Integrated optical neural network (4×2 photonic ONN) that performs convolutional operations in silicon‑nitride waveguides with graphene modulators.
- Low‑power design targeting sub‑milliwatt consumption per frame, suitable for battery‑operated and edge devices.
- Lead‑free quantum‑dot photodetectors (InSb, InAs) extending detection to 2.5 µm wavelength.
- Compatibility with standard 19‑inch rack‑mount form factor and PyTorch API for seamless software integration.
- Fabrication processes adapted from conventional semiconductor manufacturing, enabling scalable production.