Beta Release

• Create model for detecting human posture. (Standing, Sitting Lying)
  • Collect data for each category
    • Created reports for each collection session Session 1 Session 2
  • Adapt previous model to handle multiple categories instead of binary results
  • Format incoming collected data (target points (y, z, velocity, acceleration), 3x highest points, 3x lowest points)
    • Data Handling Report
• Flash produced model onto radar boards in order to assert validity using CCS (Code Composer Studio) testing.
  • Using TVM compiler to assemble flashable code
• Create visualiser to enable the viewing of collected radar data immediately
• Add CI and CD capabilites to the model workflow
  • Move some function outside of the notebooks in order to improve unit testing
  • Release generated model on each merge

Beta Release File Structure

2024-MMWAVERADARSENSORS
├── LICENSE
├── README.md
├── docs/
└── model
    ├── dry-wet/                                                     # Directory containing Dry-wet model project files
    ├── grass-notgrass/                                              # Directory containing grass-notgrass model project files
    │
    ├── posture                                                      # Directory containing posture model project files
    │   ├── DataHandlingReport.md                                    # Document describing way data is handling for training
    │   ├── dataCollectionArchive/                                   # Directory containing all data collected for posture model and reports on each collection session. Data is not necessarily collected under the same conditions and will have varying usefulness for generating a model. 
    │   ├── dataset/                                                 # Directory containg the datasets used for training the current model
    │   ├── modules                                                  # Directory containing helper functions for the jupyter notebook
    │   │   └── helper_functions.py                                  # Helper functions for jupyter notebook
    │   ├── pytorch_fp_model-POSTURE.ipynb                           # Jupyter notebook containing python for generating model from training data
    │   └── v_human-pose                                             # Directory containing python files for interacting with radar board
    │       ├── L6432_Parsing.py                                     # Python file for collecting data over UART from the board when collecting training data
    │       ├── L6432_send_cfg.py                                    # Python file for sending the configuration to the radar board
    │       ├── TrackingClassification_MidBw-POSE-115200.cfg         # Configuration file for the board when collecting data
    │       ├── motion_and_presence_detection_demo.release.appimage  # App Image to be flashed to the board when collecting training data
    │       ├── parseCommon.py                                       # Helper functions for parsing frame data recieved from the board
    │       ├── parseFrame.py                                        # "
    │       ├── parseTLVs.py                                         # "
    │       └── requirements.txt                                     # Requirements for the parsing python modules
    └── requirements.txt                                             # Requirements for the model python modules