12,000+ labeled images across 15 distinct human activity classes. The dataset includes training and validation subsets where each image is assigned a single activity label based on its directory location.
Use Cases
- Train a classification model using the folder names as ground truth labels for the 15 activity classes
- Validate model accuracy using the images contained within the validation folders
- Perform feature extraction on the 12,000+ image files to identify key human poses associated with each activity label
Strengths
- 12,000+ labeled images across training and validation sets
- 15 distinct categories of human activities
- Single-label classification format with images organized into class-specific folders