Vision-Based Pattern Recognition Models for Intelligent Human Robot Interaction in Smart Spaces

This dataset supports the research article entitled “Vision-Based Pattern Recognition Models for Intelligent Human Robot Interaction in Smart Spaces.” The dataset was prepared to document the experimental structure, preprocessing outcomes, training configuration, model evaluation, ablation analysis, and comparative performance results used in the study. The dataset contains structured sheets covering four human–robot interaction tasks: gesture recognition, object detection, activity recognition, and intention prediction. It includes dataset characteristics, sample distribution before and after preprocessing, train-validation-test splits, augmentation details, training parameters, multi-task model performance, ablation study results, baseline comparison, simulated smart space validation records, and aggregate performance summaries. The study evaluates a hybrid CNN-Transformer framework integrated with reinforcement learning for intelligent human–robot interaction in smart spaces. The dataset reflects the reported experimental results, including accuracy, precision, recall, F1-score, latency, and comparative performance against CNN-only and Vision Transformer baselines. The dataset is intended to support transparency, reproducibility, and further analysis of vision-based pattern recognition models for smart spaces, assistive robotics, healthcare environments, collaborative workplaces, and smart city applications. This dataset may be used for academic research, methodological comparison, educational purposes, and further development of intelligent robotic perception and interaction frameworks. Users are encouraged to cite the related article when using this dataset.