/waterquality

Primary LanguageJupyter Notebook

waterquality

Required:

python>=3.6
pytorch>=1.2
Keras=2.2
tensorflow=1.14

Data available online here , in npy format.

Index:

image-image7/: raw image folder (deprecated; please using .npy file for analysis). name: XXX nb_batch a/b.jpg XXX means content value, nb_batch means batch number of that image. (total nb of batch equals to nb of folders). a/b means testing under bottle a or b

classic_classification/, classic_regression/: results showing. subfolder "Gray" have some fun images

classification_result/: conventional methods comparsion experiment, in .py file

interval_result/: results by interval experiments

modelWeights/: h5 files of nn weight

result/: csv format, comparsion table, for different model settings

result_analyse/: measure how many predicted points in each bin(interval)

water/: playground for the data and model middle layer. (feat_analysis.ipynb)

website/: oil prediction website.

pytorch-cifar/:

main.py: wrapper for training models/: conventional models and new models (mainly attnResNet50.py)

/bin_analyse.py

comparsion experiments. such as test_different_pretrain_model_with_gen()

/classic_classification.py /classic_regerssion.py /classification.py comparsion experiments with different conventional methods. treated as classification / regression

/CNN_regression.py /CNN_regression2.py /CNN.py old models

/combined_trick_model.py pretrained resnet18 with linear fc

/data_loader.py load image dataset,image hight/width...

/data_to_npy.py save the processed data as .npy,

/draw_width.py draw predicted bin width

/feature_test.ipynb feature engineering, middle layer visualization, histogram analyse

/interval_process.py calculate interval array and store

/model_eval.py comparsion test different pretrain model

/multiNets.py /multiNetsVgg.py old model used in server, plain DNN/VGG

/ordinal_categorical_crossentropy.py loss function in Keras, for ordinal loss

/plot_result_fig.py scatter plot for prediction results

/predictioin_tocsv.py show_roc_pr_curve, store comparsion result in a .csv

Please cite

@misc{waterquality,
  author = {minoriwww},
  title = {waterquality project & OilSS},
  year = {2019},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/minoriwww/waterquality}}
}