Malaysian Car Plate Dataset
Description
Project Context This dataset was developed as part of a Final Year Project (FYP) at the Faculty of Information Science and Technology (FIST), Multimedia University (MMU). The broader project focuses on computational intelligence and smart campus infrastructure, specifically involving the conversion of standard local webcams into network-accessible IP cameras for real-time video surveillance and processing. This dataset serves as the foundational data to train a computer vision model for Lisence Plate Recognition (LRP), aiming to automate vehicle access control and parking management within a campus or residential environment. Content The dataset contains a carefully curated collection of images featuring Malaysian vehicle license plates. Malaysian car plates are unique because they often feature varying font styles (including italic or cursive on older/custom plates), different sizes, and both single-row and double-row formats. Total Images: Insert Number 500+ images Format: .jpg and .png Annotations: Bounding box annotations are provided in YOLO format, making it ready for plug-and-play training with modern object detection algorithms. Target Classes: 1. car_plate 2. car Collection Methodology I collected data, but some of the images are using an internet dataset, which I sorted myself. The images capture vehicles from various angles (front and rear) and under different environmental/lighting conditions (daytime glare, shadows, etc.) to ensure the trained model is robust in real-world scenarios. The images were manually reviewed, and precise bounding boxes were drawn around the license plates using Roboflow. Inspiration & Potential Use Cases This dataset is ideal for computer vision researchers and developers working on: ANPR / ALPR Systems: Training object recognition models (like YOLOv8, SSD, or Faster R-CNN) to detect and crop Malaysian license plates for downstream Optical Character Recognition (OCR). Smart Parking: Integrating real-time computer vision with IoT devices for automated boom-gate access and vehicle logging. Feature Extraction Algorithms: Testing image processing techniques (like OpenCV morphological operations, ORB, or SIFT) to isolate plate characters against complex vehicle backgrounds. Acknowledgements If you utilize this dataset in your own research, algorithms, or academic projects, please consider citing this repository. Special thanks to the Faculty of Information Science and Technology (FIST) at MMU for the academic support throughout this Final Year Project.
Files
Steps to reproduce
About this directory This dataset is structured in the standard YOLO format, making it ready for immediate use in training object detection models. The root directory is split into two main subsets: train and valid. Directory Structure: train: Contains the majority of the dataset 80% used to train the model. valid: Contains a smaller subset 20% used to evaluate the model's performance and tune hyperparameters during the training process. Inside each of the above directories: images: Contains the raw Malaysian car plate images in .jpg and .png formats. labels: Contains the corresponding .txt annotation files. Each text file shares the exact same name as its corresponding image. The text files contain the YOLO format coordinates (Class, X_center, Y_center, Width, Height) normalized between 0 and 1.
Institutions
- Multimedia UniversityMelaka, Malacca