Ultra-high Energy Efficiency AI Accelerator for Portable Healthcare Device
Description of Invention
Early screening for retinal diseases is crucial for patients with diabetes, hypertension, and retinal vascular occlusions. Regular examinations of retinal vascular status can promptly detect changes in the condition and guide treatment. However, rural residents and individuals with mobility challenges face numerous obstacles when traveling to large hospitals frequently. To address this issue, we have developed a portable AI retinal disease screening device. This device integrates local high-efficiency AI computing power, enabling data collection and professional screening to be completed without relying on the network. It rapidly provides diagnostic reports while comprehensively protecting patient privacy. By using our proprietary AI chip toolchain, we co-designed chip architecture and retinal screening algorithms to swiftly produce high-efficiency AI chips for portable devices. Our system prototype and software toolchain have notably cut product integration costs. Make retinal health checks more convenient, safe, and efficient!
Key Technology Edges
- Co-designing application algorithms with AI chip architecture boosts energy efficiency by 7x over edge-side GPUs, reducing development cycles by 50%.
- Innovative AI chip architecture cuts off-chip memory access by over 80%, offering nearly 10x energy efficiency compared to similar chips.
- Enhanced hardware prototype system, coupled with a mature proprietary software toolchain, reduces system integration and application deployment costs by 70%.