Runs object detection on a Raspberry Pi 3 using input from an attached Pi Camera. Implementation in Python using OpenCV2 is based on a MobileNet-SSD v2 model in TensorFlows ProtoBuf format.
Description of how to access Pi Camera from Python see Picamera Documentation. The explanation of how to perform inference in OpenCV2 on a TensorFlow model can be found in the OpenCV Wiki.
Install TensorFlow on Raspberry Pi through pip.
sudo apt update
sudo apt install python3-dev python3-pip
sudo apt install libatlas-base-dev # required for numpy
sudo pip3 install -U virtualenv # system-wide installvirtualenv --system-site-packages -p python3 ./venv
source ./venv/bin/activate
pip install --upgrade pip
pip install --upgrade tensorflow
# Test TensorFlow
python -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000,1000])))"The installation of OpenCV2 described in here seem 58A4 ed to have worked too.
sudo apt install libtiff5-dev libjasper-dev libpng12-dev
pip install pillow jupyter matplotlib cython
pip install opencv-pythonDownload MobileNet-SSD v2 weights and config.
-
python camera.pycamera output directly into frame buffer (fullscreen, without image capture and further processing). -
python cv_camera.pycamera image is captured into a buffer and drawn inside a X window usingcv2.imshow. No further processing of the image. -
python ssd_camera.pycapture image from camera, performan object detection, draw recognized objects with labels into image and display image in an X window.