bwiec / cis589_proj

cis589_proj

Hardware

• AMD Kria KV260
• Logitech C920 webcam

Install dependencies

Start from Ubuntu 22.04 on KV260

sudo apt-get install -y \
  python3-boto3

Install Kria PYNQ DPU

git clone https://github.com/Xilinx/Kria-PYNQ.git
cd Kria-PYNQ
sudo bash ./install.sh -b KV260; # This will take about 25 minutes

Then login as root

sudo su

and complete the installation as root

. /home/ubuntu/Kria-PYNQ/pynq/sdbuild/packages/xrt/xrt_setup.sh
. /etc/profile.d/pynq_venv.sh
pip3 install pynq-dpu --no-build-isolation

then copy the xmodel file TODO!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

AWS Setup

1. Must have an AWS IAM user with AmazonRekognition* permission policies (add a group with all those policies then add user to group)
2. Set up target environment to communicate with aws

mkdir -p ~/.aws
touch ~/.aws/credentials
touch ~/.aws/config

where ~/.aws/credentials looks like

[default]
aws_access_key_id = your_access_key_id
aws_secret_access_key = your_secret_access_key

and ~/.aws/config looks like

[default]
region = your_aws_region

see https://docs.aws.amazon.com/pdfs/rekognition/latest/dg/rekognition-dg.pdf#what-is

Environment setup

source /home/ubuntu/Kria-PYNQ/pynq/sdbuild/packages/xrt/xrt_setup.sh
source /etc/profile.d/pynq_venv.sh

Run the cloud app

./cloud.py -h

For example, to detect faces using an image on disk

./cloud.py --print_duration --display --test_image ./irishterrier-696543.jpg --algorithm detect_faces

or to run on the live camera

./cloud.py --display

Run the edge app

1. Login as root

sudo su

2. As root, setup the environment

. /home/ubuntu/Kria-PYNQ/pynq/sdbuild/packages/xrt/xrt_setup.sh
. /etc/profile.d/pynq_venv.sh
xauth add $(xauth -f ~ubuntu/.Xauthority list | tail -1)

3. Run the script

./edge.py -h

For example, to run the algorithm using an image on disk, print information, and display the result

./edge.py --print_duration --display --test_image ./irishterrier-696543.jpg

or to run on the live camera

./edge.py --display