get_data.py crawls data and json from OpenI
json_to_csv.py create labels from labels.csv
resize_pic.py resize picture to 224*224
shuffle_origin.py reads data and labels from picture and csv(contains 6738 items), shuffles them and save as train.npy and eval.npy
read_array.py contains functions reads data and labels from train.npy and eval.npy
nin.py contains model and train routes
Accuracy 60.55% Precision 67.02% Recall 73.58% F1score 70.03%
loss&f1score: