pics = np.genfromtxt(imgs_loc, delimiter=',')
boxs = np.genfromtxt(boxs_loc, delimiter=',')
# Resize the box lists
boxs = boxs.reshape(boxs.shape[0], -1, 4)
# dimensions of the whole pictures, constant for now
picshape = 240, 240 # height and width
# Create the NN model
localizer = Localizer(picshape=(240, 240), hidden_layers=(42, 30, 10))
epochs = 20
batch_size = 20
localizer.train(pics, boxs, epochs, batch_size)
i = random.randint(0, pics.shape[0] - 1)
pic = pics[i, :].reshape(240, 240)
utils.test_localizer(localizer, pic, show_probs=True)
#print(localizer.predict(pic))
# a hard test, just for fun
resistors = Image.open('resistors.png', mode='r')
resistors = resistors.convert(mode='F')
utils.test_localizer(localizer, np.asarray(resistors), show_probs=True)
localizer.save('datasets/dataset{0}/best_model'.format(dataset))
# Create the NN model
localizer = Localizer(input_shape=(48, 48), hidden_layers=(42, 30, 10))
# Create the data generator
generator = PictureGenerator(batch_size=15,
batches_per_epoch=500,
return_angles=False,
resistor_prob=0.5,
real_backgrounds=True,
angle_num=8,
flatten=False)
# Train the model
epochs = 10
localizer.train(generator, epochs)
#i = random.randint(0, pics.shape[0] - 1)
#pic = pics[i,:].reshape(240,240)
#utils.test_localizer(localizer, pic, show_probs=False)
#resistors = Image.open('test_pictures/hard_test.png', mode='r')
#resistors = resistors.convert(mode='F')
#
#utils.test_localizer(localizer, np.asarray(resistors), show_probs=False)
#localizer.save('datasets/dataset{0}/best_model'.format(dataset))
