Using our SDK and ML experience in the Embedded field we are now optimizing popular networks and preparing them to run on your devices based on Arm and Xilinx platforms. Currently these work on local simulators and we are working to provide them on real hardware and Arm Virtual Hardware (Corstone-300 U55/U65 in the cloud).
If you are developing end devices we are able to provide you ready-to-run networks as well as develop new ones to fit your scenarios.
If you are developing neural networks we can help you adapt and optimize them for your target platforms.
Optimized neural network models that we have made to run on Arm Cortex-M55 and Ethos-U55
- Body parts detection (MoveNet.Thunder) – head, shoulders, elbows, knees
- Supersampling (Deep Convolution Network)
- … we are regularly working to add new items to this list
Â
Optimized neural network models that we have made to run on Xilinx Kria System-on-Modules (SoMs)
- Face landmarks detection (PFLD)
- Face recognition (Densebox)
- Supersampling (Deep Convolution Network)
- … we are regularly working to add new items to this list