Buoy Object Detection with TensorFlow's Mobilenet model

Object detection allows for the recognition, detection, and localization of multiple buoys within an image using a live video feed

Installation

Download and Install Docker

sudo apt-get update
sudo apt-get install \
    apt-transport-https \
    ca-certificates \
    curl \
    software-properties-common
curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add -
sudo add-apt-repository \
   "deb [arch=amd64] https://download.docker.com/linux/ubuntu \
   $(lsb_release -cs) \
   stable"
sudo apt-get update
sudo apt-get install docker-ce
sudo groupadd docker
sudo usermod -aG docker $USER

Build the Dockerfile to install all dependencies

Note: If the following docker commands do not work, run it with sudo (or log out and log back in).

Note: This must be run in the root folder of this repository
Alternatively: Replace . with /path/to/Dockerfile

docker build -t  tf-buoy-classifier .

Verify that the image has been successfully built using

$ docker images

REPOSITORY              TAG                 IMAGE ID            CREATED             SIZE
tf-buoy-classifier      latest              272ef1a20710        10 seconds ago      2.54GB
tensorflow/tensorflow   1.4.0-py3           7d680bfcec87        4 months ago        1.25GB

Make sure you are in the root directory of this repository and start the docker container with:

xhost +
docker run -it --rm --privileged -p 8888:8888 --env DISPLAY=$DISPLAY -v /dev/video0:/dev/video0 --volume="/tmp/.X11-unix:/tmp/.X11-unix:rw" -v $(pwd):/home/TF tf-buoy-classifier:latest

Usage

The following must be run inside the docker container

Build and install python

If running for the first time, run:

python3 -B setup.py build
python3 -B setup.py install

python3 -B object_detection/object_detection_runner.py

Based of

MIT LICENSE

Realtime-Object-Detection