matheusyukio / face-recog-dl Goto Github PK

Split functions and files

` train_data_generator = dataTrainAugmentation().flow_from_dataframe(
dataframe=training_data,
directory = os.path.join(os.getcwd(),'../input/lfw-dataset/lfw-deepfunneled/lfw-deepfunneled/'),
target_size = (250, 250),
x_col = "image_path", y_col = "name",
batch_size=BATCH_SIZE,
class_mode = "categorical",
shuffle = True)

    X, y = train_data_generator.next()
    X_valid, y_valid = valid_data_generator.next()
    print("===========Train============")
    print(X.shape)
    print(y.shape)

`
Using GPU the training process with Matrix is faster than ImageGenerator but the model overtiffing using always the same X and y from .next()
Create a function to iterate and feed model.fit

Add Dockerfile and add README about how to run using Docker

todo: https://www.tensorflow.org/addons/tutorials/losses_triplet
model = tf.keras.Sequential([ tf.keras.layers.Conv2D(filters=64, kernel_size=2, padding='same', activation='relu', input_shape=(28,28,1)), tf.keras.layers.MaxPooling2D(pool_size=2), tf.keras.layers.Dropout(0.3), tf.keras.layers.Conv2D(filters=32, kernel_size=2, padding='same', activation='relu'), tf.keras.layers.MaxPooling2D(pool_size=2), tf.keras.layers.Dropout(0.3), tf.keras.layers.Flatten(), tf.keras.layers.Dense(256, activation=None), # No activation on final dense layer tf.keras.layers.Lambda(lambda x: tf.math.l2_normalize(x, axis=1)) # L2 normalize embeddings])

model.compile( optimizer=tf.keras.optimizers.Adam(0.001), loss=tfa.losses.TripletSemiHardLoss())