longchanging / magnetsensor

using deep learning model with magnetism data to classify different platform and applications.

MagnetSensor

using deep learning model with magnetism data to classify different platform and applications.

Author

Lanqing Yang,SJTU [email protected] [email protected]

Hao Pan,STTU

Folder

Source code version description

V0.0 LSTM&baseline

src V0.0 Initial LSTM&baseline

mainly other_classifier.py    (√) 
config.py                     (√)     
LSTM.py                       (√)
Model.py                      (√)
preprocess.py                 (√)
read_data.py                  (√)

other_classifier.py     use other models to do classification
config.py     important paras for LSTM and baseline
LSTM.py       main LSTM function to classify
read_data.py  read data and simply preprocessing(fft,feature engineering)
preprocess    PCA and min-max preprocessing
Model.py      construct LSTM model using tf

V1.0

src V1.0 FCN Model

# FCN folder 
config.py           (√)
mydata_model.py     (√)
predict.py          (√)
prepare_data.py     (√)

# LSTM folder 
mainly other_classifier.py   
config.py     
LSTM.py           
Model.py          
preprocess.py     
read_data.py       

config.py          config just of FCN
mydata_model.py    main FCN model
predict.py         use FCN to predict
prepare_data.py    prepare data for FCN


# LSTM folder      use them to generate data for FCN and baseline

V2.0

src V2.0 feature Extraction

config.py
feature_core.py          (√)
feature_fft.py           (√)
feature_reExtraction.py  (√)
feature_time.py          (√)
train_test.py            (√)

This folder is forked here

https://www.zhihu.com/question/41068341?sort=created 

Not yet completed,just a trial.

V2.1

src V2.1 data explore,try plot raw data

plot.py                  (√)
sample_test.py           (√)
test_smooth.py           (√)

Failure trial,try plot normal raw data.
To explore more about min-max model,and others.
Finally achieved in V3.1  

V3.0

src V3.0 Add main,setup,argv,combine fcn&baseline,call together

Note: From this version, folder structure is quiet different.

baseline.py   
config.py
main.py             (√)
plot.py
prepare.py
preprocess.py
read_data.py 
setup.py            (√)
train_test.py

main.py         Combine all functions together
		Entrance of all
		argv included to get user command like 'train/test/predict' 

setup.py     Generate configs for config.py 

V3.1

src V3.1 Changed package name for convenience

o0_DataExploer.py  mainly to explore the data and plot some pdfs.

o0_DataExploer.py   (√)
o0_dataExplore.py   (√)
o0_plot.py          (√)
o1_setup.py
o2_config.py
o3_read_data.py
o4_preprocess.py
o5_prepare.py
o6_train_test.py
o7_baseline.py
o8_main.py