| language: | |
| - en | |
| license: mit | |
| library_name: keras | |
| tags: | |
| - human-activity-recognition | |
| - pose-estimation | |
| - mediapipe | |
| - tensorflow | |
| datasets: | |
| - guillherms/human-activity-pose_v4 | |
| pipeline_tag: image-classification | |
| pretty_name: Human Activity Recognition MLP | |
| # 🧠 Human Activity Recognition MLP | |
| This model classifies human activities (reading, waving, office work, etc.) | |
| using pose landmarks extracted from **MediaPipe Pose**. | |
| ## 📊 Architecture | |
| - Multi-Layer Perceptron (Dense NN) | |
| - Input: 132 pose landmark features | |
| - Output: 10 activity classes | |
| ## 🧩 Files | |
| - `mlp_pose.h5`: Trained Keras model | |
| - `label_encoder.pkl`: Encodes activity labels | |
| ## 🧠 Author | |
| Created by **Guilherme Santos** | |