Indian License Plates with Labels

The “Indian License Plates with Labels” dataset is a comprehensive collection focusing on Indian vehicle number plates. Initially, the dataset consisted of 181 images, meticulously downloaded and manually annotated using the makesense.ai tool. This foundational effort laid the groundwork for a much larger and diverse dataset, expanding significantly to encompass thousands of images from various platforms and environments. These images include a wide range of vehicle types such as cars, motorcycles, trucks, and buses, captured in different lighting conditions, angles, and backgrounds, ensuring robustness and versatility for various machine learning applications.

To further enhance the dataset, a Yolov5 object detection model was applied, enabling the incorporation and labeling of nearly 2000 additional unlabeled images from a secondary source. This expansion significantly increased the dataset’s utility, making it suitable for a broad spectrum of research and development activities in automatic number plate recognition (ANPR).

The dataset is organized into two main folders: one for images and another for labels. The labels are in YOLO format, which details the coordinates, width, and height of the license plate boundary boxes. This organization facilitates ease of use and integration with various machine learning frameworks and tools, allowing researchers and developers to efficiently train and test their models.

The diversity in vehicle types and the variety of capture conditions included in the dataset ensure that models trained on this data can generalize well to real-world scenarios. This makes the dataset particularly valuable for developing robust ANPR systems capable of performing accurately under various conditions.