#1 Vision AI with Arty Z7 for Person Detection
Table of Contents
1 Description of Person Detection with yolo
Embedded vision is having a transformative effect on skill sets, technical innovation, and product development. There are many technical solution to embedded vision, such as High Performance MCU, MPU, NPU. AMD adaptive SoCs and FPGAs are special in this domain. It enrich largest flexibility and performance for developers. It can perform the vision tasks continuously with expectable latency and nearly lowest power consumption. Utilizing Arty Z7 kit based on the AMD Zynq 7000 SoC devices, no cooling fan is needed.
One of entry level Vision AI tasks is person detection. That is what proposed in this project. Person detection is widely used in now days. Many city bus equipped with person detection and person counting technology, that is how we can find the vacancy of each bus from apps like google map bus transition suggestion.
There are many AI model for person detection, yolo is one of the efficient one. It appears on Model Zoo by AMD and it seems that it can give and easy start point for this project. There are still many high-end model under-developing, such as InsightFace: an open source 2D&3D deep face analysis library, it offers instance face detection for multi-persons.
For other vision models, refer to InsightFace: an open source 2D&3D deep face analysis library.
2 Vision AI with Arty Z7
The Arty Z7 is a ready-to-use development platform designed around the Zynq-7000 All Programmable System-on-Chip (AP SoC) from Xilinx, integrated a dual-core, 650 MHz ARM Cortex-A9 processor with Xilinx 7-series Field Programmable Gate Array (FPGA) logic. With rich peripheral, USB, Ethernet, HDMI, Audio, Pmod Connectors Arduino/chipKIT shield connectorj,2 slide switches,4 Push buttons and 4 LEDs plus 2 RGB LEDs.
The Vivado, Petalinux, and Vitis AI both make software solution, support kinds of programming with whatever BareMetal, digital logic, even with linux environment. Furthermore, one PYNQ project provide full-function programming with python-only virtual environment.
3 Arty Z7 and USB camera
Here is delivery of the sponsored development kit from element14.
First, Arty Z7-20, Arty Z7 - Digilent Reference,
Second, ZIQIAN HD 1080p Webcam with tripod,
Third, Pmod ALS: Ambient Light Sensor , buy from 410-286 DIGILENT, 附加板, PmodALS, 单环境光传感器, 8位, 转换光至数字日期
Forth, MULTICOMP PRO MP001996, buy from MP001996 MULTICOMP PRO, 交流/直流电源, 6级, ITE, 1输出, 24 W, 12 VDC, 2 A
Besides these main parts, there are other accessary parts ready for this project.
4 Unboxing with out-of-factory Image
Plug the power adaptor and power on Arty Z7 board, the out-of-factory image runs, blinking leds and being responsive to push-button and slide buttons. The oob Image is stored in SPI memory, it keeps running until memory is erased. Normal codes can online debug on RAM, it disappear after power off.
This proves that the board is in good condition.
5 Challenges and Opportunity
While the challenges is great even for this person detection AI model. Before this first blog, I have tried different ways to run the AI model and still can not pass them all. Before this challenge, I have used Pynq-z1 development board with pathfinding algorithm based on pynq-dpu. It seems new version of pynq-dpu do not support this board directly.
To Complete the project in time poses challenges. Good news is I have chances to dive deep on how it works. Next blogs shall follow the path on DPU and how to make it work.
