- Download and unzip datasets.
- Confirm annotations included the start and stop frames at the beginning of the
.txt
. - Excludes dataset without annotations.
- Run
fallDetectorV0\preprocess.py
.- this creates the MHIs and the contour feature csv files
- PCA mean and eigenvalues are saved in
fallDetectorV0\pcaEig.joblib
- Run
fallDetectorV0\training.py
- sklearn SVM object is saved in
fallDetectorV0\svm.joblib
- sklearn SVM object is saved in
- Run
fallDetectorV0\training.ipynb
.- Weights are saved to
trained_model\fdnet.pt
- Weights are saved to
- Run
fallDetectorV0\onlineFallDetector.py
Not implemented
Note: Code reference for MHI and CNN training: https://medium.com/diving-in-deep/fall-detection-with-pytorch-b4f19be71e80