Human Action Recognition HAR Dataset
Image dataset representing 15 different classes of human actions (e.g. running, sitting, dancing), with approximately 12,000 images for training and 5,400 for testing.
Approximately 18,000 JPEG images divided into folders by classes
Open Database License (ODbL) probable — check the terms of use
Description
The dataset Human Action Recognition HAR includes images classified according to 15 common human actions. Each image belongs to a single category, stored in specific folders, making it easy to build CNN models for action recognition.
What is this dataset for?
- Training models for the classification of human actions from images
- Develop monitoring systems or behavioral analysis
- Testing CNN architectures adapted to visual recognition
Can it be enriched or improved?
Yes, this dataset can be completed with video annotations, multimodal data (skeleton, depth), or enriched with contextual metadata. The diversity of poses can also be increased by data augmentation.
🔎 In summary
🧠 Recommended for
- Computer vision students
- HAR researchers
- AI developers
🔧 Compatible tools
- TensorFlow
- PyTorch
- OpenCV
- Keras
💡 Tip
Use image augmentation techniques to improve the robustness of models.
Frequently Asked Questions
What is the exact nature of the dataset license?
The dataset is probably under Open Database License (ODbL), it is advisable to check the precise conditions of use on the source.
How many human action classes does this dataset contain?
15 different classes, covering various daily activities.
Does this dataset include videos?
No, only classified static images.
