Multi-label classification

test2.py

import os, sys, cv2
import numpy as np
from data_manager import DataManager
from model import get_model, get_cust_model

n_classes = 5
shape = (128, 128, 3)

def train(inp_path):
    trn = DataManager(inp_path, shape, 'TRAIN')
    val = DataManager(inp_path, shape, 'VALID')

    model = get_cust_model(shape, n_classes)

    model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
    model.fit(trn.dataset[0], trn.dataset[1], epochs=10, validation_data=(val.dataset[0], val.dataset[1]), batch_size=64)
    model.save('scenery_10.hd5')
    model.evaluate(x=val.dataset[0], y=val.dataset[1], batch_size=64)

def test(inp_path, model_path):
    model = get_cust_model(shape, n_classes)
    model.load_weights(model_path)
    #val = DataManager(inp_path, shape, 'VALID')
    val = DataManager(inp_path, shape, 'TRAIN')
    corr, size = 0, len(val.dataset[0])
    for i in range(len(val.dataset[0])):
        image, labels = val.dataset[0][i], val.dataset[1][i]
        image = np.expand_dims(image, axis=0)
        predn = model.predict(image)
        predn = np.array([0 if e < 0.46 else 1 for e in predn[0]])
        if np.equal(predn, labels).all():
            corr += 1
    print ('acc = %.2f' % float(corr/size))

if __name__=='__main__':
    if sys.argv[-1] == '-train':
        train(sys.argv[1])
    else:
        test(sys.argv[1], sys.argv[2])