ANU_YOLO

A collection of YOLO-based scripts and environment setups for face and object detection, designed for use on Raspberry Pi OS and other Linux systems.

---

Table of Contents

• Project Overview
• Prerequisites
• Installation
• Model Weights
• Usage
• Group Task
• Credits

---

Project Overview

This repository provides scripts for running YOLOv8-based face and object detection on images and videos. The main scripts, object.py and yolo_face.py, detects objects and faces using a YOLOv8n model and its retrained model and outputs annotated images or videos with bounding boxes and confidence scores, respectively. This introduces the newcomer to various ML concepts.

Preparation

Simply copy and paste this onto the terminal and run it. If everything runs correctly the system will auto-reboot and you can reopen the terminal and everything should work including pip directly as well as python. This preperation is immaterial of whether you have cloned, forked etc.

#!/usr/bin/env bash
set -euo pipefail
IFS=$'\n\t'

# 1) System update & upgrade
echo "1) Updating system…"
sudo apt-get update
sudo DEBIAN_FRONTEND=noninteractive apt-get upgrade -y

# 2) Repair any broken packages
echo "2) Repairing broken packages…"
sudo apt-get install -f -y
sudo dpkg --configure -a

# 3) Install prerequisites
echo "3) Installing prerequisites…"
sudo DEBIAN_FRONTEND=noninteractive apt-get install -y \
    curl wget unzip bzip2 build-essential cmake \
    libjpeg-dev libpng-dev libtiff-dev \
    libavcodec-dev libavformat-dev libswscale-dev ffmpeg \
    git ca-certificates

# 4) Clean up
echo "4) Cleaning up…"
sudo apt-get autoremove -y
sudo apt-get autoclean -y

# 5) Keyboard layout → US
echo "5) Setting keyboard layout to US…"
sudo sed -i 's/^XKBLAYOUT=.*/XKBLAYOUT="us"/' /etc/default/keyboard
sudo DEBIAN_FRONTEND=noninteractive dpkg-reconfigure -fnoninteractive keyboard-configuration
sudo udevadm trigger --subsystem-match=input --action=change

# 6) Python tooling & Ultralytics YOLO (no venv)
echo "6) Installing Python tooling & YOLO…"
sudo apt-get install -y python3 python3-pip libatlas-base-dev libopenblas-dev
python3 -m pip install --upgrade --user pip >/dev/null
pip install --user numpy ultralytics requests >/dev/null

# 7) Ensure ~/.local/bin is on PATH
PROFILE="$HOME/.bashrc"
LINE='export PATH="$HOME/.local/bin:$PATH"'
if ! grep -Fxq "$LINE" "$PROFILE"; then
  echo "$LINE" >> "$PROFILE"
fi

# 8) Verify pip functionality by installing and importing numpy
echo "8) Verifying pip can install & import numpy…"
if python3 -c "import numpy" &> /dev/null; then
  echo "pip check passed: rebooting now…"
  sudo reboot
else
  echo "numpy not found, attempting install…" >&2
  if pip install --user opencv-python >/dev/null && python3 -c "import opencv-python" &> /dev/null; then
    echo "pip check passed after install: rebooting now…"
    sudo reboot
  else
    echo "ERROR: pip failed to install or import numpy." >&2
    exit 1
  fi
fi

Error that I made was upgrading your systems but most of you did not complete properly, so it might have broken things. You should not need conda or any other packages/containers after simply installing and changing paths.

For those who have not cloned or forked, fork this repo and then run:

git clone https://github.com/<your_username>/ANU_YOLO.git
cd ANU_YOLO

For those who have already forked the repo and need to update it to the current state of this repo:

# 1. Change directory to your local repository
cd ANU_YOLO

# 2. Add the original repository as an upstream remote (do this only once)
git remote add upstream https://github.com/1ssb/ANU_YOLO.git

# 3. Fetch changes from the upstream repo
git fetch upstream

# 4. Checkout your main branch (replace 'main' with your branch name if different)
git checkout main

# 5. Merge changes from upstream/main into your local main branch
git merge upstream/main

# 6. Push the updated main branch to your fork on GitHub
git push origin main

---

Model Weights

Download the YOLOv8 face detector weights from the official repository:

• lindevs/yolov8-face

Place the downloaded weights file (e.g., yolov8n-face-lindevs.pt) in the project root directory.

---

Usage

Object Detection

python object.py --source bus.jpg --output bus_objects.jpg

Face Detection on Images

python yolo_face.py --source bus.jpg --output bus_face.jpg

• Input: bus.jpg (or any image file)
• Output: bus_face.jpg (annotated image with bounding boxes and confidence values)

Face Detection on Videos

python yolo_face.py --source path/to/video.mp4 --output path/to/output.mp4

• Input: Any video file
• Output: Annotated video with detected faces

Streamer

python stream.py

This should open the webcam interface if you attach one for livestreaming the detection.

---

Group Task

Count the number of apples in the video: apples.mp4. Submit your script as apples.py (one per team). Please list your team name and members here:

...

Credits for video: https://www.istockphoto.com/video/gala-apple-falling-gm472985251-20196243?utm_source=pixabay&utm_medium=affiliate&utm_campaign=sponsored_video&utm_content=srp_topbanner_media&utm_term=apple+fall

---

Credits

• YOLOv8 by Ultralytics: https://docs.ultralytics.com/models/yolov8/
• Face model weights: https://github.com/lindevs/yolov8-face
• Josh Starmer: Videos --> I went to LSE to study Machine Learning in 2019 only to end up watching his videos for the exams.
• ChatGPT for a faux annotated image used in the presentation.