English
You are here:
AI on the Edge Accelerator
We offer the first Mixed Quantization digital IP core on the market with native 8bit CNN and 1bit BNN support. We develop our own open-source training framework to develop Binary Neural Networks on top of the standard Tensorflow/Keras stack.
Features
- Scalable configuration from 8 to 128 PE
- Synthesizable IP core
- Power- & performance-optimized configurations
- External RAM option
- Standalone & accelerator configurations
Deliverables
- HDL code and testbenches
- Reference firmware and models
- ASIC and FPGA constraints
- Documentation
- Reference implementations
Applications
- Audio Wake Word (AWW)
- Video Wake Word (VWW)
- Human Presence Detection
- Gesture Recognition
- Crowd Detection
View AI on the Edge Accelerator full description to...
- see the entire AI on the Edge Accelerator datasheet
- get in contact with AI on the Edge Accelerator Supplier