object-detection-with-raspberry-pi-5

Object detection with Raspberry Pi 5, Raspberry Pi Camera Module 3, and Google Coral Edge TPU

How To Setup M.2 or Mini PCIe Accelerator Coral Edge TPU with a Virtual Environment

Introduction

This README guide provides step-by-step instructions on setting up the Coral Edge TPU, either in the M.2 or Mini PCIe form-factor. We will also cover setting up a Python virtual environment for running TensorFlow Lite models on the Edge TPU.

Requirements

• A computer with one of the following operating systems:
  • Linux: 64-bit Debian 10 or Ubuntu 16.04 (or newer), x86-64 or ARMv8 architecture
  • Windows: 64-bit Windows 10, x86-64 architecture
• Support for MSI-X as defined in the PCI 3.0 specification
• At least one available Mini PCIe or M.2 module slot
• Python 3.6-3.9

Setup Steps

1. Connect the Module

• Ensure the host system is shut down.
• Connect the Coral Mini PCIe or M.2 module to the corresponding slot on your host system.

2. Install the PCIe Driver and Edge TPU Runtime

On Linux

• Check your Linux kernel version: uname -r
  • If it's 4.18 or lower, proceed with the installation.
  • If it's 4.19 or higher, check for a pre-built Apex driver: lsmod | grep apex
    • If nothing is printed, proceed with the installation.
    • If an Apex module is listed, follow the workaround to disable Apex and Gasket.
• Add the Debian package repository:

echo "deb https://packages.cloud.google.com/apt coral-edgetpu-stable main" | sudo tee /etc/apt/sources.list.d/coral-edgetpu.list
curl https://packages.cloud.google.com/apt/doc/apt-key.gpg | sudo apt-key add -
sudo apt-get update

• Install the PCIe driver and Edge TPU runtime:

  sudo apt-get install gasket-dkms libedgetpu1-std

• If necessary, add a udev rule and ensure your user is in the 'apex' group:

sudo sh -c "echo 'SUBSYSTEM==\"apex\", MODE=\"0660\", GROUP=\"apex\"' >> /etc/udev/rules.d/65-apex.rules"
sudo groupadd apex
sudo adduser $USER apex

• Reboot the system and verify the module is detected: lspci -nn | grep 089a
• Verify the driver: ls /dev/apex_0

On Windows

• Install the latest Microsoft Visual C++ 2019 redistributable.
• Download and extract edgetpu_runtime_20221024.zip, then run install.bat.

3. Install the PyCoral Library

On Linux

sudo apt-get install python3-pycoral

• If you receive an installation error that looks like the below snippet, you will have to setup a Python virtual environment with pyenv with the Python version being between 3.6 and 3.9 (see Step 4):

The following packages have unmet dependencies:
 python3-pycoral : Depends: python3-tflite-runtime (= 2.5.0.post1) but it is not going to be installed
                   Depends: python3 (< 3.10) but 3.11.2-1+b1 is to be installed
E: Unable to correct problems, you have held broken packages.
pi@pi:~$

On Windows

python3 -m pip install --extra-index-url https://google-coral.github.io/py-repo/ pycoral~=2.0

4. Setup Python Virtual Environment with pyenv

• Add pyenv to ~/.bash_profile:

export PYENV_ROOT="$HOME/.pyenv"
export PATH="$PYENV_ROOT/bin:$PATH"
eval "$(pyenv init -)"
eval "$(pyenv virtualenv-init -)"
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/lib

• Create and activate a new virtual environment:

cd /path/to/your/project # Where you want to create the virtual environment, install packages, and run your code, e.g.: cd ~/Documents/Projects/Repos/object-detection-with-raspberry-pi-5
sudo apt-get install -y python3-venv
pyenv install 3.9
pyenv virtualenv 3.9 myenv
pyenv activate myenv

• Upgrade pip and install necessary packages:

pip install --upgrade pip
pip install pycoral Pillow numpy

5. Run Pre-Built Example Model on the Edge TPU

Navigate to the example directory and run a classification model

mkdir coral && cd coral
git clone https://github.com/google-coral/pycoral.git
cd pycoral

Download the model, labels, and bird photo

bash examples/install_requirements.sh classify_image.py

Run the image classifier with the bird photo

python3 examples/classify_image.py \
--model test_data/mobilenet_v2_1.0_224_inat_bird_quant_edgetpu.tflite \
--labels test_data/inat_bird_labels.txt \
--input test_data/parrot.jpg

Troubleshooting on Linux

• Resolve HIB errors on ARM64 platforms by modifying kernel arguments to include gasket.dma_bit_mask=32.
• Resolve pcieport errors by adding pcie_aspm=off to the kernel command line arguments.
• To disable pre-built Apex and Gasket drivers, use the /etc/modprobe.d/blacklist-apex.conf workaround.

Next Steps

• Explore other models and example projects provided by Coral.
• Train your own models using TensorFlow and make them compatible with the Edge TPU.

For more detailed information, visit the Coral Accelerator Get Started Guide.

Resources

• Install Docker Engine on Debian
• Install TensorFlow Lite for Python on a Raspberry Pi
• TensorFlow Lite runtime package - Quickstart for Linux-based devices with Python