A classic implementation of Federated Learning (FedAvg) for identifying FALL
and ADL
from images utilizing Transfer Learning Tensorflow and Numpy.
- VGG16 as Feature Extractor
- Federated Averaging
Data Source - UR Fall Detection Dataset
The dataset contains 70 videos illustrating FALL
and ADL
events. We've extracted RGB frames from the videos for this work and augmented the dataset by employig common augmentation techiniques (e.g. rotation, flipping, etc.).