Skip to content

Build CNN_LSTM Recognition Models on eSense DataSet using TensorFlow

License

Notifications You must be signed in to change notification settings

teco-kit/CNN_LSTM

 
 

Repository files navigation

CNN_LSTM

How to get the data?

DataProcess_raw2vec.py

  • First step of the project, convert raw data into the form that the following steps need.
  • Build 2 folders './Vecs' and './Seps' before running this

Before Training

  • After running DataProcess_raw2vec.py, build a folder './Models/Saved_Model'
  • Build 2 folders './Seps/train' and './Seps/test', divide files in './Seps' randomly into these two folders

Training

  • There are 5 models provided for comparation
  • CNNLSTM_cnn.py: Vanilla CNN Model
  • CNNLSTM_lstm.py: Vanilla LSTM Model
  • CNNLSTM_fsw.py: Intraframe-CNN-LSTM Model
  • CNNLSTM_osw.py: Interframe-CNN-LSTM Model
  • CNNLSTM_nsw.py: Combi-CNN-LSTM Model
  • The trained models will be saved in './Models/Saved_Model'
  • For definations of the last 3 models, please refer to my master thesis "An Earable Interactive System Based on Head Motion Recognition from the Data Captured from IMUs" for more details

ResultFilter.py

  • Including postprocessing steps (details in my thesis) and evaluation

ModelTransform.py

  • Transform Tensorflow-Models into Tflite-Models, so that you can easily deploy this into any Android APP.
  • Build a folder './tflite_models' before running this

quaternion.py, vector.py

  • Including functions related to Quaternion Calculation and Vector Calculation

About

Build CNN_LSTM Recognition Models on eSense DataSet using TensorFlow

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%