whsair / Wearable-Multimodal-Optical-and-Acoustic-Sensing-of-Physiological-Parameters-Autoencoder-

Wearable Multimodal Optical and Acoustic Sensing of Physiological Parameters (Autoencoder)

Aurther: Hongshuo Wang

Director: Professor Sri-Rajasekhar (Raj) Kothapalli & PHD. SUMIT AGRAWAL

Range: 05/2021 - 08/2021

Only includes code

Installed packages

• numpy
• scipy
• pdb
• pyusb
• sklearn
• matplotlib
• tflite_runtime (during running)
• tensorflow 2.x (training)

Instruction on lab Pi

How to check the storage of the pi:

df

How to run the trained tensorflow-lite model:

1. open the command windows

2. type cmd (go to the tflite1 directory):

cd /home/pi/tflite1

3. type cmd (activate python visual-envir):

source tflite-env/bin/activate 

4. type cmd: jupyter notebook (open jupyter notebook)
   • demo2.ipynb (short demo to show the final workout)
   • load_lite_model_new.py (testing tensorflow lite model on testing set)
   • measure_running_speed.ipynb (compare normal tensorflow model vs tensorflow lite model in running speed)
   • Plotter_test.py (live demo on input streaming signal)
   • X_test.csv (X value)
   • y_test.csv (y value)
   • converted_model.tflite (converted tensorflow lite model)
   • my_model_80_5000.h5 (og version of tensorflow model)

How to display live data(running Plotter_test.py or Plotter_test_og)

1. open the command windows

2. type cmd (go to the tflite1 directory):

cd /home/pi/tflite1 

3. type cmd:

sudo thonny Plotter_test_og.py
sudo thonny Plotter_test.py

4. click run(green) icon

How to train the tensorflow model using given X and y

• run training_data.ipynb

How to convert h5 trained weight to tensorflow-lite version

• run generate_tf_lite_model.py

How to compare the results between lite model and og model

• run measure_running_speed.ipynb