LeapMind is currently developing "Efficiera", an AI inference accelerator IP with ultra low power consumption, optimized for CNN inference processing on FPGA or ASIC devices.
Efficiera enables integration of AI features into various edge devices such as home appliances, industrial machines such as construction machinery, surveillance cameras, broadcasting devices, and small machines and robots that are limited in power, cost and heat dissipation, which were previously technically challenging.
- Energy Efficient
- Efficiera reduces the power required for convolution operations by minimizing the amount of data transfer and the number of bits.
- High Performance
- The reduction of the arithmetic logic reduces the number of arithmetic cycles and also improves the arithmetic performance per area and frequency.
- Small footprint
- The reduction of the arithmetic logic reduces the area of the silicon and minimizes the area per arithmetic unit while maintaining performance.