def test_transform_image_dataframe_to_matrix(): min_images_per_person = [5] min_per_person = min_images_per_person[0] multi_data = get_mounted_data(min_per_person, min_per_person) #Y = multi_data[['name']] #X = multi_data[['image_path']] #print(Y.head(6)) #print(X.head(6)) data_x, data_y = transform_image_dataframe_to_matrix( multi_data, 250, 250, 'lfw-dataset/lfw-deepfunneled/lfw-deepfunneled/')
def test_create_dataset_tfRecord(): min_images_per_person = [100] min_per_person = min_images_per_person[0] multi_data = get_mounted_data(min_per_person, min_per_person) #Y = multi_data[['name']] #X = multi_data[['image_path']] #print(Y.head(6)) #print(X.head(6)) create_dataset_tfRecord(multi_data, 'lfw-dataset/lfw-deepfunneled/lfw-deepfunneled/')
def main(): min_images_per_person = [30] models = ["DeepFace"] num_folds = 5 batch_sizes = [60] for min_per_person in min_images_per_person: for batch in batch_sizes: for model in models: multi_data = get_mounted_data(min_per_person, min_per_person) Y = multi_data[['name']] X = multi_data[['image_path']] CLASSES = Y.groupby('name').nunique().shape[0] print("### run_k_fold ", " min_per_person ", min_per_person, " CLASSES ", CLASSES, "model ", model, " batch_size ", batch) run_k_fold(multi_data, X, Y, CLASSES, model, batch, num_folds) tf.keras.backend.clear_session() gc.collect()
def main(): epoch = 10 min_images_per_person = [130]#[30,25] # [25,20] models = ["LeNet5"]#,"VGGFace"]#,"AlexNet","DeepFace","VGGFace"] num_folds = 2 #aumentando o batch para 30 DeepFace conseguiu bons resultados, testar com outras batch_sizes = [30]#[2,4,8,30] for min_per_person in min_images_per_person: for batch in batch_sizes: for model in models: multi_data = get_mounted_data(min_per_person, min_per_person) Y = multi_data[['name']] X = multi_data[['image_path']] CLASSES = Y.groupby('name').nunique().shape[0] # print("### run_hold_out "," epoch ", epoch, " min_per_person ", min_per_person," CLASSES ", CLASSES,"model ",model," batch_size ",batch) # run_hold_out(multi_data, X, Y, CLASSES, epoch, model, batch) print("### run_k_fold ", " epoch ", epoch, " min_per_person ", min_per_person, " CLASSES ", CLASSES, " model ", model, " batch_size ", batch) run_k_fold(multi_data, X, Y, CLASSES, epoch, model, batch, num_folds)
def main(): epoch = 300 min_images_per_person = [30] #[30,25] # [25,20] models = [ "AlexNet", "DeepFace" ] #["LeNet5","AlexNet","DeepFace"]#["LeNet5","AlexNet","DeepFace","VGGFace"] num_folds = 5 batch_sizes = [30, 60] #[2,4,8,30] for min_per_person in min_images_per_person: for batch in batch_sizes: for model in models: multi_data = get_mounted_data(min_per_person, min_per_person) Y = multi_data[['name']] X = multi_data[['image_path']] CLASSES = Y.groupby('name').nunique().shape[0] print("### run_k_fold ", " epoch ", epoch, " min_per_person ", min_per_person, " CLASSES ", CLASSES, "model ", model, " batch_size ", batch) run_k_fold(multi_data, X, Y, CLASSES, epoch, model, batch, num_folds) gc.collect()
def main(): min_images_per_person = [30] models = ["DeepFace"] num_folds = 5 batch_sizes = [60] min_per_person = min_images_per_person[0] multi_data = get_mounted_data(min_per_person, min_per_person) Y = multi_data[['name']] X = multi_data[['image_path']] CLASSES = Y.groupby('name').nunique().shape[0] nomes_classes = [] for i in pd.DataFrame( Y.groupby('name')['name'].nunique().reset_index( name="unique"))['name']: nomes_classes.append(str(i)) print('CLASSES') print(CLASSES) #print(nomes_classes) train, test = train_test_split(multi_data, test_size=0.25, random_state=42, shuffle=True, stratify=multi_data[['name']]) print('train') print(train.shape) print(train) print(train.groupby('name').count().image_path) print('test') print(test.shape) print(test) print(test.groupby('name').count().image_path) calfw_df = calfw_mount_data(nomes_classes) print(calfw_df.shape) calfw_Y = calfw_df[['name']] calfw_X = calfw_df[['image_path']] calfw_CLASSES = calfw_Y.groupby('name').nunique().shape[0] #print(calfw_df.groupby('name')['name'].nunique().reset_index(name="unique")['name']) print(calfw_Y) print(calfw_X) print(calfw_CLASSES) ind_counts = calfw_df.groupby('name').count().image_path print(ind_counts) calfw_classes = [] for j in pd.DataFrame( calfw_df.groupby('name')['name'].nunique().reset_index( name="unique"))['name']: calfw_classes.append(str(j)) print(calfw_classes) print(set(nomes_classes) - set(calfw_classes)) print(set(calfw_classes) - set(nomes_classes)) for batch in batch_sizes: for model in models: print("### run_k_fold ", " min_per_person ", min_per_person, " CLASSES ", CLASSES, "model ", model, " batch_size ", batch) run_k_fold(calfw_df, X, Y, CLASSES, model, batch, num_folds, nomes_classes) tf.keras.backend.clear_session() gc.collect()