guillherms/

human-activity-pose_v2

	🧍 Human Activity Pose Dataset


This dataset contains pose landmarks extracted using **MediaPipe Pose**,  
annotated with **activity labels** and English descriptions.

## 📊 Dataset Structure
- `train/` → 80% of the samples for training  
- `validation/` → 20% of the samples for validation

## 🧠 Each record contains:
- 33 pose keypoints (`x, y, z, visibility`)
- `label`: activity name (e.g., `reading`, `dancing`, `office_work`)
-… See the full description on the dataset page: https://huggingface.co/datasets/guillherms/human-activity-pose_v2.