IoTKITs: A Novel Dataset for IoT Education Kit Recognition

Overview

IoTKITs is a curated and well-annotated image dataset introduced in Data in Brief in 2025 to support recognition and classification of IoT education kits such as Arduino, ESP32, and Raspberry Pi boards. It focuses on full education kits rather than just PCB components.

Dataset Composition

• 3,200 RGB images of IoT education kits.
• 32 kit classes, including Arduino Uno, Arduino Nano, ESP32, Raspberry Pi variants, and other commonly used education kits.
• Images captured in real-world lab or classroom environments as well as on plain backgrounds.
• Most images contain a single kit; a subset includes multiple boards per image for added variability.

Data Collection and Annotation

• Images collected from multiple sources, including Kaggle, Roboflow Universe, and original photographs by the authors.
• All images manually annotated using Roboflow Smart Polygon, creating polygon masks that tightly follow kit contours.
• Annotations provided in JSON format compatible with common object detection and instance segmentation frameworks.
• Dataset split with roughly 80 training and 20 testing images per class, yielding a balanced distribution.

Intended Use and Baselines

• Designed for object detection and classification of IoT education kits in educational and embedded systems contexts.
• Baseline experiments evaluate detectors such as YOLOv5, YOLOv7, Faster R-CNN, and SSD on IoTKITs.
• Useful for training lightweight CNN-based models for deployment in educational tools and lab monitoring systems.

Access and License

The Data in Brief article describes IoTKITs and links to the dataset, which is publicly available via associated repositories (e.g., GitHub or Zenodo). Users should follow the citation instructions in the article and respect any license terms stated on the hosting page.