Computer Vision Model for Object Detection for AROS

This project is designed to detect specific classes using a TensorFlow Lite model (MobileNet V2 640x640).

Classes detected

• Obstacles
• person
• bicycle
• car
• motorcycle
• bus
• truck
• accident

Metrics

Sample of Real Time Detection

Watch the video

Steps

Step 1: Generate TFLite Files

Walk through the TFOD tutorial up to step 12 to generate TFLite files.

Step 2: Clone the Repository

Clone the current repository onto your Raspberry Pi or copy it from a machine using RDP.

git clone https://github.com/OmarZakaria10/Road-Obstacles-ObjectDetection-Tflite.git

Step 3: Set Up Virtual Environment

Create and activate a virtual environment to manage project dependencies.

On Unix or MacOS:

python3 -m venv venv
source venv/bin/activate

On Windows:

python -m venv venv
venv\Scripts\activate

###Step 4: Install Dependencies Install the required dependencies listed in the requirements.txt file

pip install -r requirements.txt

Additionally, install the following system dependencies on your Raspberry Pi:

sudo apt-get install libcblas-dev libhdf5-dev libhdf5-serial-dev libatlas-base-dev libjasper-dev libqtgui4 libqt4-test
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

Step 5: Run Real-Time Detections

python3 script.py