Automated Detection of Military Vehicles from Video Input (ADOMVI)

Introduction

This repository contains notebooks and resources used to train a state-of-the-art military vehicle tracker. Its main focus is on building a dataset of relevant images and annotations to fine-tune pre-trained object detection models, namely a Yolov8 model.

We start by building a training dataset from images available in open source object detection datasets (ImageNet, OpenImages, Roboflow). We also use scraping tools to collect more images of military vehicles from Google images. This allows us to define four broad classes of military vehicles that our model can then discriminate: Armoured Fighting Vehicle (AFV), Armoured Personnel Carrier (APC), Military Engineering Vehicle (MEV) and Light Armoured Vehicle (LAV). We provide a sample annotated dataset for these classes.

We also explore using diffusion models and the dreambooth method to generate new training images in different scenes and conditions.

Contents

• The adomvi directory contains utility functions to fetch and format datasets for training a Yolov8 model for object detection.
• The resources directory contains video samples for vehicle detection task.
• The notebooks directory contains exemple notebooks on how to
  1. Prepare a custom dataset of images annotated for automatic detection of military vehicles.
  2. Train train a Yolov8 model using the prepared dataset.
  3. Run tracking using the trained model on a sample video.
  4. Fine tune Dreambooth to generate images of a tank.

Installation

To install the project, clone the repository and install the project in a python environment, either using pip

git clone [email protected]:jonasrenault/adomvi.git
cd adomvi
pip install --editable .

or using poetry

git clone [email protected]:jonasrenault/adomvi.git
cd adomvi
poetry install

Run the notebooks

To run the notebooks, start a jupyter lab server with

jupyter lab

and open one of the notebooks in the notebooks directory.

Tracking of military vehicles with multi-class object detection model

Some sample results of tracking different types of military vehicles (AFV, APC, MEV, LAV) using a finetuned yolov8-large model.

Generating diversity in our training dataset using Stable Diffusion and dreambooth