
Plumerai develops embedded AI accelerator IP core for FPGAs
London-based embedded machine learning specialist Plumerai has developed an IP core for FPGAs using its proprietary low power Binarised Neural Network (BNN).
The Ikva accelerator runs BNNs and also efficiently supports 8-bit models using Plumerai’s tool flow and ultra-fast and memory-efficient inference engine integrated with TensorFlow Lite. A 32-bit RISC-V processor controls Ikva, captures the data from the camera and provides a programmer-friendly runtime environment.
- World’s fastest deep learning inference software for ARM Cortex-M
- XMOS and Plumerai partner on binarised neural networks
“During the development of Ikva, we aimed to design a new hardware architecture for our optimized AI models while keeping it highly flexible and suitable for unknown future models,” said the company. “In contrast to other AI companies that seem to either develop models, or training software, or AI processors, we focus on the full AI stack and the Ikva core completes our offering. With Ikva, we now support the full AI stack starting from data collection, to training and model development, to very efficient inference engines, and now all the way down to providing the most optimized hardware implementations.
Ikva fits in small and low-power FPGAs such as the Lattice CrossLink-NX but is scalable, both in memory and in compute power.
Plumerai has ported one of its proprietary person presence detection models together with the inference software running on the Ikva IP core in a Lattice CrossLink-NX LIFCL-40 FPGA. This is a low-power and low-cost 6x6mm FPGA that is available off-the-shelf and includes a native MIPI camera interface, further reducing the number of components in the system.
Ikva runs the person presence detection model 10x faster on the CrossLink-NX FPGA than on a typical Arm Cortex-M microcontroller. Alternatively, the frame rate can be scaled down to 1 or 2fps for those applications where low energy consumption is key.
This could be used in the home to automatically turn off the TV, lights, or heating when there’s no one in the room. Outside the home, the doorbell can send a message when there’s someone walking up to the front door or a small camera can detect when there’s an unexpected visitor in the backyard. In the office, the PC can automatically lock the screen when not in use, saving battery life.
Other ML models also run on the Ikva core, which is available today with the supporting tool flow,and optimized person detection models.
Related articles
- Nordic ports TinyML to cellular IoT chip
- TinyML platform builds AI models in minutes
- MicroSys, Hailo team for embedded AI platform
- AI-enabled embedded controllers for always-connected IoT
Other articles on eeNews Europe
- ARM details virtual model use for the IoT
- UK signs quantum technology deal with US
- Rohde & Schwarz re-structures as turnover drops
- European tech combines LoRa and cellular for smart farming
- Analog Devices shows digital energy tech in Europe
- Europe to build independent SiC value chain
