2.Requirements
- python >= 3.6
- keras >= 2.2.4
- scikit-learn >= 0.19.1
- opencv-python >= 3.2.0.6
- gensim >= 3.6.0
- pandas
- numpy
- scipy
- PIL
3.Modules
tool_utils.py :for all data processing;
tool_wordembeddings.py:loading trained word-embeddings of label-list,source：https://github.com/xgli/word2vec-api
model_AE.py:auto-encoder for class-wordembeddings;
model_VGG.py:image features extraction；
model_PSN.py:Pseudo-Siamese Network(PSN),using cosine-distance of class-wordembeddings as output label;
model_SAE.py:Semantic Auto-encoder(SAE);
model_CVAE.py:Conditional Variational Auto-encoders(CVAE);
model_AttClassifiers.py:Attributes classifiers;
tool_distance.py:Input std_we and prediction_we,calculating cosine distances;
model_CS.py:Calibration stacking algorithm(CS),input 'predict-cosine-distances.csv',output 'submit.csv';
tool_testing.py:for testing.