Accelerator equipment for neural network video analytics on Intel FPGA
Equipment development and production of the Russian Federation
Support for Intel OpenVinoToolKit: standard video analytics development toolkit on neural networks, ready-made library of pre-trained neural network models
Implementation of neural network analytics for the “Safe City” program
We focus on the implementation of neural network video analytics for the Safe City program.
We have developed and serially produce 2 types of equipment for video analytics tasks
- Embedded SOM video camera (54x44mm)
- PCIe-accelerator embedded in the video server (168x69x28mm)
New Accelerator EulerProject HPC with support for Intel Open VINO Toolkit
- Form factor ½ PCI Express 168 mm x 69 mm x 28 mm
- Host interface 8-lane PCI-Express Gen 3
- Two DDR4 controllers (HPC version), each bank of 8-16GB
- Intel Arria-10 FPGA 20nm
How to use Euler HPC with OpenVINO support
The new accelerator of the Russian development and production of EulerProject HPC based on Intel® Arria 10 1150GX PCIe FPGA can be easily integrated into a small server.
IntelOpenVinotoolkit support implemented
- Mainstream N.S. support frameworks (Caffe, TensorFlow)
- Opportunity optimizations
- The possibility of developing custom neural network convolution topologies or modifications of primitives N.S.
Intel OpenVINO library of pre-trained models
The library is actively developing and expanding. The following models are currently available:
- Generic Age & Gender Recognition
- Camera Tampering Detection
- Generic Face Detection
- Face Detection for Retail
- Person Detection for Retail
- Face Detection for Automotive
- Person, Vehicle & Bike Detection
- Vehicle License Plate Detection
- License Plate Recognition
- Age & Gender Recognition for Retail
- Vehicle Attributes Recognition
- Head Pose Estimation for Automotive
- Semantic Segmentation
- Road Segmentation
- Person Attributes
- Person Re-identification
- Pedestrian Detection
- Pedestrian & Vehicle Detection
- Emotions Recognition