| | --- |
| | license: apache-2.0 |
| | task_categories: |
| | - text-classification |
| | - table-question-answering |
| | language: |
| | - en |
| | tags: |
| | - agent |
| | - code |
| | size_categories: |
| | - 1K<n<10K |
| | --- |
| | |
| | Dataset Overview:- |
| |
|
| | This is a high-fidelity synthetic dataset designed for training AI models in Human-Robot Interaction (HRI) and Privacy-Preserving Home Monitoring. Unlike standard vision datasets, this file provides 4D Point Clouds (Spatial + Velocity) specifically tuned for 24GHz mmWave Radar sensors. |
| | The dataset features complex "Long-Tail" scenarios, such as children running in low-light environments, specifically modeled to include real-world physical noise and sensor artifacts. |
| | 🛠 Technical Specifications |
| | Sensor Simulation: 24GHz FMCW mmWave Radar |
| | Format: JSONL (JSON Lines) |
| | Environment: Low-light/Night-time Living Room |
| | Subjects: 2 Humanoids (including Child_5yo_Running class) |
| | Engine: Anode_Noise_v1 (Synthetic Physics Engine) |
| | 🧬 Data Features |
| | Each data point is enriched with high-dimensional physics metadata that standard datasets lack: |
| | Kinematic Vectors: Cartesian coordinates (x, y, z) + Velocity Vectors (vx, vy, vz). |
| | Surface Physics: Includes surface_normal and stress_strain_tensor for advanced collision modeling. |
| | Semantic Intelligence: Frame-by-frame labels for pose (e.g., short_stride_sprint), intent_prediction (e.g., crossing_room), and intent_confidence. |
| | Safety Metrics: Real-time threat_vector and closing_speed_mps calculations for autonomous braking/slowing systems. |
| | Ethical Priority Tags: Integrated ethical_priority labels (e.g., "monitor + slow_if_close") for training Responsible AI. |
| |
|
| | Licensing:- |
| | |
| | This is a Commercial-Tier Dataset. For full access to the 1M+ frame sequence or the underlying Anode Physics Engine, please contact: |
| | vesperbyarservice@gmail.com |
