Nigerian Native Headwear Detection: Project Overview

---

The purpose of this project was to use Object Detection (Computer Vision) detect some of the most popular traditional headgears worn in Nigeria namely;

• Abeti Aja from the South West,
• Awolowo from the South West,
• Fila from the South West,
• Gele from the South West;
• Fula from the North;
• Okpu-Ozo from the South-East
• Okpu-Agu from the South-East
• the Turban

---

• Used Google Web Scraper to collect the data/images

• Performed data-label annonatation using Label Studio

• Used YoloV5 to train Object Detection model on the dataset

---

Model Performamce

Confusion Matrix

F1_curve

PR_curve

---

How to Detect Images

• Clone repo and install requirements.txt in a Python>=3.7.0 environment, including PyTorch>=1.7.

  git clone https://github.com/EyimofeP/nigerian-headwear  # clone
  cd yolov5
  pip install -r requirements.txt  # install

• Add the images or videos you want to detect in the detect-img/ folder

• Run this command in your terminal

    python detect.py --weights best.pt --source detect-img/img-name.jpg 
                                                detect-img/img-video.mp4
                                                'https://youtu.be/Zgi9g1ksQHc'  # YouTube
                                                0 #Webcam

• Go to the runs/detect/ folder to see the results!