author：Luo Ya'nan
-
caffe_try
Classification use caffe's python API.
!!!not correct yet.
- findTheSameData: Use for validating correction.
- readLMDB: Transfrom data from lmdb to numpy.
- save_feature: Save fc's feature maps and predicitions.
- validate_param: Validate whether the paramerters of the deploy and train_test prototxt's net are the same.
-
mnist
Try mnist data set use caffe's python API, get correct result.
- load_mnist_data: Transform mnist data into the visiable form.
- myload_mnist_data: Change some of the load data function, 具体什么忘记了...太久没用了
- predict_label: Use API to batch classify test data.
- mnist_solver: Train net hyperparams.
- mnist_train_test: Train net model.
- mnist_deploy: Prediction use, which is not include data layer.
- mnist_mnist: Train net use Caffe command.
-
tf_try
Classification use Tensorflow.
-
matlab_plot
- plotting: Plorring curves.
- read_test: Read test accuracy, loss, etc.
- read_train: Read train accuracy, loss, etc.
-
python_analyze
- center_or_border_statistics: Get center and border samples statistics
- data_analysis: Get data mean, std, etc. And plot corresponding curves.
- draw_test: Visualization of test samples' classification.
- is_Generator: Judge whether GAN get good training.
- plotting: Plotting accuracy and loss curves verses iterations.
- read_data: Get the data after train and test.
- t_sne: Dimensionality reduction.
-
GANs
- cgan_tfonly.py: Conditional GAN.
- gans_config: GAN configrations.
- generator: generator of GANs.
- preprocess_data: Preprocessing data and save .mat for each class.
- train_gans: train GANs.
-
data_preprocessing: Extract and divide the original data set into train and test data set according to ratio.
-
data_preprocess_pos: Data preprocessing increases save position information.
-
load_data_tricks: Add class-0 for train.
-
generate_gaussian_noise: Generate gaussian noise.
-
test: Validate whether the data extract is correct with BPN net model.
-
deep_cnn: Deep net model.
-
original_cnn: Original net model.
-
train_original: Train original method.
-
train_deep: Train deep net method.
-
cnxgboost: Use cnn-fc output as feature to train xgboost.
-
train_test: 没有在做了，不完整的代码......
-