HandGesture-5G: A Multi-View, Multi-Illumination Smartphone Image Dataset of Five Static Hand Gestures for Robust Computer Vision Research
Description
This dataset contains 2,653 high-resolution color images of five static hand gestures: Open Palm (Open Hand), Fist, Pointing Up, L-Sign, and V-Sign. The images were acquired using Apple iPhone 14 Pro Max and Apple iPhone 15 Pro Max smartphones, with several acquisition sessions stabilized using a DJI Osmo Mobile 3-axis gimbal. For each gesture class, both the palmar (front) and dorsal (back) views of the hand were captured under diverse indoor illumination conditions, ranging from low-light (underexposed) to bright (overexposed) environments, thereby increasing the variability and robustness of the dataset for computer vision applications. The dataset was collected from seven volunteers (five male and two female participants) in Laboratory 302, Department of Computer Science and Engineering, Jahangirnagar University, Savar, Dhaka, Bangladesh. Images were captured under controlled yet naturally varying lighting conditions to support the development and evaluation of robust hand gesture recognition systems. The dataset is organized into five class-specific folders containing 661 Open Palm, 542 Fist, 405 Pointing Up, 537 L-Sign, and 508 V-Sign images. Most images are stored at their original smartphone resolutions (6048 × 6048 and 3024 × 3024 pixels) in JPEG format, with a small number of HEIC images. Each image has been characterized according to its mean grayscale luminance and categorized into Low, Correct, or High exposure levels to facilitate illumination-aware computer vision research. This dataset was primarily created as part of the sub-project "Developing a Gesture-based Intelligent Wheelchair for the Physically Challenged People in Bangladesh" under the Improving Computer and Software Engineering Tertiary Education Project (ICSETEP), funded by the Asian Development Bank (ADB) and the University Grants Commission of Bangladesh (UGC). The dataset will be used in future research to develop and evaluate vision-based wheelchair navigation and human–wheelchair interaction systems based on static hand gesture recognition. The dataset supports research in computer vision, hand gesture recognition, human–computer interaction (HCI), artificial intelligence, deep learning, and assistive technologies. It can be used for model training, benchmarking, transfer learning, illumination-robust recognition, view-invariant learning, explainable AI (XAI), and image classification tasks. The images contain only hands and do not include faces or other directly identifying personal information.
Files
Steps to reproduce
1. Download and extract the dataset. 2. The dataset is organized into five folders corresponding to the gesture classes: Open Palm, Fist, Pointing Up, L-Sign, and V-Sign. 3. Read the images in their original JPEG (and a small number of HEIC) formats using any standard image processing library (e.g., OpenCV, Pillow, or TensorFlow/PyTorch data loaders). 4. Resize images to the desired input resolution (e.g., 224 × 224 pixels) if required by the selected deep learning model. 5. Split the dataset into training, validation, and testing subsets according to the experimental design (e.g., 70/15/15 or 80/10/10). 6. Apply optional preprocessing techniques such as normalization, data augmentation, or exposure correction depending on the research objective. 7. Train and evaluate computer vision or deep learning models for tasks including hand gesture recognition, image classification, human–computer interaction, view-invariant learning, and illumination-robust recognition. 8. Compare model performance using standard evaluation metrics such as Accuracy, Precision, Recall, F1-score, and Confusion Matrix.
Institutions
- Jahangirnagar UniversityDhaka Division, Dhaka