Intelligent Control (ECE5759) Final Term Project: Object Detection in Aerial Image

Overview

Object detection

Object detection is the task of detecting instances of objects of a certain class within an image. Each detected objects are usually marked with a bounding boxes along with a confidence score and a class ID.

Here is a demo:

YOLOv5

YOLOv5 🚀 is a family of object detection architectures and models pretrained on the COCO dataset, and represents Ultralytics open-source research into future vision AI methods, incorporating lessons learned and best practices evolved over thousands of hours of research and development.

VisDrone

• Drones, or general UAVs, equipped with cameras have been fast deployed to a wide range of applications, including agricultural, aerial photography, fast delivery, and surveillance.
• Automatic understanding of visual data collected from these platforms become highly demanding, which brings computer vision to drones more and more closely.

Tasks & Goals

Tasks

• In this homework, you will learn how to use YOLO object detection models on real-world surveillance datasets.
• The model, data, training and testing pipelines have been provided.
• Your job is to perform training and testing YOLOv5 models on the given dataset.
• You can adjust some basic parameters to make the training faster, or to obtain higher accuracy.

Goals

• Given a limited amount of training time (2 weeks), you must obtain the highest accuracy on the given dataset.
• You must make a submission file of the test-set:
  • Run prediction scripts (in the Jupyter Notebook) and obtain the results (.txt files) for each image.
  • Compress the labels folder (containing all .txt files) and rename it to: student_id.zip.
  • Upload the .zip file to CodaLab website.

Scoring Rule

• We will use the popular COCO mAP@50 score to judge the accuracy of object detection models.
• More details about this metric can be found here: link
• Final score = Your mAP / 0.3539 * 100

Submission link: https://codalab.lisn.upsaclay.fr/competitions/4806

• Register using school email account (No exception)
• Max total submissions: 100
• Max submissions per day: 10

Deadline: 11:59 PM on May 30, 2022

Getting Started

Prerequisite

• OS: Ubuntu 20.04 / 22.04.
• Environment: Python>=3.9.0 and PyTorch>=1.11.0 with anaconda (Recommend).
• Editor: Jupyter Notebook.

To getting started, you must install anaconda before hand. Then open the terminal and get started:

cd <to-wherever-you-want-to-store-the-project-files>

# Create a new environment named `project`
conda create --name project python=3.9

# Activate the environment (you must activate the conda environment everytime you start the terminal
conda activate project

# Install Jupyter Notebook and clone the `project` repo. 
# You only need to call this command once. If you have already install it, just skip this step
conda install git jupyter
git clone https://github.com/SKKU-AutoLab-VSW/ece5759_final_term_project
cd ece5759_final_term_project
pip install -r requirements.txt

# Finally, run Jupyter Notebook and open the `final_project.ipynb` file in your web browser:
jupyter-notebook