Avnet Ultra96-V2 Face Detection Tutorial

The Avnet Ultra96-V2 single-board computer (SBC) enables hardware and software developers to explore the capabilities of the Zynq UltraScale+ MPSoC. To begin, a 16GB microSD card which holds the boot image and necessary demo files has been provided as the boot device.

microSD Card image - Ultra96-V2: http://avnet.me/ultra96v2-face-detect-demo-image

We'll use the densebox model and deploy it on the Ultra96-V2 in a face detection application. In the following 4 steps, we will successfully run the face detection application:

1. microSD card installation instructions using Etcher

2. Connecting everything

3. Booting instructions

4. Running the face detection application

So, let’s begin!

Step 1) microSD Card installation instructions using Etcher

i. Download the Ultra96 microSD Card image, unzip/untar your download and be sure to recall the location of the file.

ii. Download Etcher SD Card installation tool for your host machine.

iii. Insert microSD card into host machine (Minimum requirement of 16GB)

iv. Follow the Etcher prompt:

• Select the unzipped/untared image from your file directory
• Select drive to which you wish to write (i.e. you microSD card)
• Click “Flash!” Button
• Wait for microSD card to completely flash and follow prompt.

v. Removed microSD card from host machine.

Step 2) Connecting everything

i. Insert the microSD card with the Ultra96 Image in the microSD card slot of the Ultra96-V2 Single Board Computer and make sure that the boot mode is set to sd_card (Ultra96: OFF ON)

ii. Connect the Ultra96-V2 to a DisplayPort monitor with a miniDP to DisplayPort adapter cable

iii. Connect the Ultra96 USB-to-JTAG/UART Pod

iv. Connect the USB webcam

v. Connect the Power Supply to the Ultra96-V2

Pro tip: Ensure to set the boot mode switch SW3 to SD boot – 1-Off, 2-On and Connect the power supply last. As soon as the power supply is connected, 12V is hot on the board, although the Pmics are disabled.

Step 3) Booting instructions

After loading the Linux image on the card, it can be up and running as follows:

i. On connecting the the power supply to the Ultra96, the green LED D17 will illuminate.

ii. Momentarily press the button labeled SW4 / POWER. It is located between the two USB 3.0 Type A ports, the green power on LED D2 will illuminate.

iii. The first sign that the boot process has begun will be a blue LED labeled D1 / DONE illuminating. This means that the bistream for the FPGA portion of Ultra96v2 has successfully loaded. This happens during uboot.

iv. Once LED D7 begins to show a regular “heartbeat” Linux is up and running.

v. From a laptop or other device search for and connect to the WI-FI network called Ultra96_xxxxxxxxxxxx, where the 12 x’s correspond to Ultra96’s MAC ID.

vi. From a browser, go to http://192.168.2.1

vii. You will be greeted by Ultra96v2’s homepage

Pro tip: Connecting your browser to the Ultra96 is sometimes disallowed by company VPN so turn VPN off if it doesn’t work.

Step 4) Running the face detection application

i. We can access the Ultra96 terminal using a PuTTY terminal application.

ii. Set COM port and Connect the Serial terminal at 115200 to the Ultra96 FPGA

iii. Obtain the running steps of ultra96 on serial terminal

iv. Now you can enter login and password details on the terminal

login : root ; password: root

v. Change directories to the face_detection application and execute the program.

vi. cd /media/card/face_detection./face_detection.elf

You can obtain more detailed instructions regarding the Ultra96-V2 face detection application by enrolling for the Ultra96 Technical Training Course (TTC).

The Ultra96 TTC comprises of 6 courses and nearly 8 days worth of training. You will learn the fundamentals of developing software applications, building a custom hardware platform, and running the PetaLinux tools to build embedded Linux for Ultra96. Along with that, these courses allow you to dive deeper into application specific areas including SDSoC, Artificial Intelligence, and Python/PYNQ.

Also, all 6 courses have instructions and solutions for either the Ultra96-V1 or the Ultra96-V2, so if anyone wants to run the Face Detection demo on their Ultra96-V1, they can get the source and solution as part of the TTC.

Enroll at https://www.hackster.io/workshops/ultra96

Done with this and ready to take it to the next level ? Try the Vitis-AI 1.0 Flow for Avnet Vitis Platforms Tutorial. Compile the DNNDK API based Applications and Vitis-AI-Library based Applications such as adas_detection, face_detection, segmentation and video_analysis.

Learn more: https://www.hackster.io/AlbertaBeef/vitis-ai-1-0-flow-for-avnet-vitis-platforms-2231f8

Thanks and stay tuned for the next one !