A brief overview of what's been done, and what's on my plate.
- Implemented second draft at recording data UI I added a 2 second timer, and a toggle to switch between recording normal data, and recording a fall.
- Added display to see num samples Normal vs Fall
- added logic for the timer
- Created fixed fixed size feature vectors
- Uploaded feature vectors to Parse
- Query Parse for num samples on update
- can now extract all data to a text file for processing
-
- Train a perceptron or simple feed forward neural network to classify
- Break up data into 80-20 Train-Test
- Compute Accuracy
- Save weights to external file
- Train a perceptron or simple feed forward neural network to classify
- Train a perceptron or simple feed forward neural network to classify
- Break up data into 80-20 Train-Test
- Compute Accuracy
- Save weights to external file
- Build Product App
- Build UI
- Clean UI
- Load classifier into App with weights from that file
- Real-time classify and notify user
- Naive notification can just be text / color change. No need for push.
- MLPNeuralNet is a pain to get running as a MacOS / Terminal App
- Compatibility issues with certain libraries being in
Objective-C
and others inSwift