#se obtienen los paths de los archivos del dataset
for feature_idx, feature in enumerate(features):
paths = list(absoluteFilePaths(baseline_path + feature))
image_set += paths
#se cargan o se generan las activaciones
if generar_activaciones:
activations = get_activations(model,
image_set,
target_layer_name,
pooled=True)
save_activations(activations, feature_type, target_layer_name,
"embedding_classification")
else:
try:
activations = load_activations(feature_type, target_layer_name,
"embedding_classification")
except Exception as e:
print("No se pudo cargar activaciones")
print("Generando nuevas activaciones")
activations = get_activations(model,
image_set,
target_layer_name,
pooled=True)
save_activations(activations, feature_type, target_layer_name,
"embedding_classification")
umap_accuracies = []
for unn in umap_nneighbors:
reducer = umap.UMAP(n_neighbors=unn)
embedding = reducer.fit_transform(activations)
image_set += paths
#se cargan o se generan las activaciones
if generar_activaciones:
activations = get_activations(model,
weights,
image_set,
target_layer_name,
pooled=True)
save_activations(
activations, feature_type + "_" + target_layer_name +
"_embedding_activations_" + weights + ".npy")
else:
try:
activations = load_activations(feature_type + "_" + target_layer_name +
"_embedding_activations_" + weights +
".npy")
except Exception as e:
print("No se pudo cargar activaciones")
print("Generando nuevas activaciones")
activations = get_activations(model,
weights,
image_set,
target_layer_name,
pooled=True)
save_activations(
activations, feature_type + "_" + target_layer_name +
"_embedding_activations_" + weights + ".npy")
targets = [x // n_ejemplos for x in range(0, n_ejemplos * len(features))]
targets = np.array(targets)
for feature_idx, feature in enumerate(features):
paths = list(absoluteFilePaths(baseline_path + feature))
feature_paths = divide_set(paths, training_fraction)
feature_training_set = feature_paths[0]
feature_validation_set = feature_paths[1]
n_training_samples = len(feature_training_set)
n_validation_samples = len(paths) - n_training_samples
training_set += feature_training_set
validation_set.append(feature_validation_set)
if generar_activaciones:
activations = get_activations(model, training_set, target_layer_name, pooled=True)
save_activations(activations, tipo_de_atributo, target_layer_name)
else:
activations = load_activations(tipo_de_atributo, target_layer_name)
#inferencia utilizando knn
series_results_by_feature = []
for feature_index, feature_set in enumerate(validation_set):
feature_results = np.zeros(len(n_neighbors))
print("calculando vecinos de atributo " + features[feature_index] + " - " + "{:.1f}".format(
feature_index * len(feature_set) * 100 / (len(feature_validation_set) * len(features))) + "%")
for file_index, file in enumerate(feature_set):
result = partial_forward(model, target_layer_name, file)
result = np.expand_dims(result, axis=0)
pooled_result = tf.keras.layers.GlobalAveragePooling2D()(result)
for idx, k in enumerate(n_neighbors):