tra_i=0
tes_i=0
datas = os.listdir('./data')
print(datas)
for e in datas:
img = cv2.imread('./data/'+e)
if e[0] == 'p':
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
x_train[tra_i] = img
y_train.append([0])
tra_i+=1
if e[0] == 'a':
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
x_train[tra_i] = img
y_train.append([1])
tra_i+=1
return (x_train,np.array(y_train)) , (x_test,np.array(y_test))
(x_train, y_train), (x_test, y_test) = load()
print(x_train.shape)
print(y_train.shape)
y_train = keras.utils.to_categorical(y_train, num_classes)
x_train /= 255
print(x_train.shape)
print(y_train.shape)
img_input = keras.layers.Input(shape=(224, 224, 3))
model = MobileNet(input_tensor=img_input, classes=num_classes)
model.summary()
model.compile(loss='categorical_crossentropy',optimizer = keras.optimizers.rmsprop(lr=0.0004, decay=5e-4),metrics=['accuracy'])
model.fit(x_train, y_train, validation_split =0.2,batch_size = batch_size, epochs = epochs, verbose = 1,shuffle=True)
convFiles[num] = (str(convFiles[num]) + f'.{extension}')
return convFiles
# Sorts images and labels into arrays
imgfiles = imgsort(os.listdir(args.img_dir), extension='png')
with open(args.label_dir, 'r') as f:
labels = f.read().split(',')
labels[len(labels) - 1] = labels[len(labels) - 1][:-1]
labels = np.expand_dims(np.array(labels), axis=1)
print(imgfiles)
print(labels)
print(labels.shape)
print("Parsing image files...")
imgs = np.array([cv2.imread(f'{args.img_dir}{img}')
for img in imgfiles]) / 255.0
print(imgs.shape)
print("Loading Model...")
model = MobileNet(classes=16)
model.compile(optimizer='adam', loss='categorical_crossentropy')
model.fit(imgs, labels, epochs=EPOCHS, batch_size=BATCH_SIZE)
outputs = model.predict(imgs[:3])
print(f"outputs:\n{[letters(i) for i in outputs]}")