Image Object Detection

Installation

This project uses pipenv as its virtualenv management tool, which you will need to have installed:

1. Install pipenv

pip install pipenv

2. You will need to make sure you have tkinter installed. This can be done as the following according to your OS:

# Check if tkinter is installed:
python -m tkinter

# Otherwise, install it

# UBUNTU / DEBIAN
sudo apt-get install python3-tk

# MacOS
brew install [email protected]

# Fedora
sudo dnf install python3-tkinter

# CentOS
sudo yum install python3-tkinter

# Make sure to specify correct Python version:
sudo apt-get install python3.10-tk

Now, just run the following command to activate the env and install all the dependencies

3. Activate the virtual environment

pipenv shell

4. Install dependencies

pipenv install

5. Downloading the models

We want to download the models and store them inside ml/models/. To achieve this, you can do the following:

cd ml
mkdir models
cd models
# Download tiny-yolov3 (size = 34 mb)
wget https://github.com/OlafenwaMoses/ImageAI/releases/download/3.0.0-pretrained/tiny-yolov3.pt 
# Download yolov3 (size = 237 mb)
wget https://github.com/OlafenwaMoses/ImageAI/releases/download/3.0.0-pretrained/yolov3.pt
# Download retina-net (size = 130 mb)
wget https://github.com/OlafenwaMoses/ImageAI/releases/download/3.0.0-pretrained/retinanet_resnet50_fpn_coco-eeacb38b.pth

# Go back to the project directory
cd ../..

Starting the server

Ensure you are in the project directory and then simply:

python -m backend.model_server

If you get an import error when running the server, you might need to do the following:

export PYTHONPATH=pwd # path to this project directory

or if you are using Windows:

set PYTHONPATH=%PYTHONPATH%;C:\project_path

Command line tool

For the command line tool, you will need to specify the path to your input image, the path & name that you want for your output image & output csv and select one from the following models: retina-net, yolov3, tiny-yolov3.

Simply run:

python cmd_interface.py detech -h 

To learn how to pass in the arguments

There are sample images present in the input_images directory and a output_images directory where you can store your results as a way to test out this project.