def predict(config):
# load json and create model
json_file = open(config['output_model_path'], 'r')
loaded_model_json = json_file.read()
json_file.close()
loaded_model = model_from_json(loaded_model_json)
# load weights into new model
loaded_model.load_weights(config['output_weight_path'])
print("Loaded model from disk")
#sgd = SGD(lr=1e-3, decay=1e-6, momentum=0.9, nesterov=True)
#loaded_model.compile(optimizer=sgd, loss='categorical_crossentropy', metrics=['accuracy'])
new_model = to_heatmap(loaded_model)
img = image.load_img('COLOR_MAP_0_C_A.png', target_size=(350, 150))
#im = preprocess_image_batch(['examples/3.png'], color_mode="bgr")
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
#x = preprocess_input(x)
out = new_model.predict(x)
heatmap0 = out[0, 0]
heatmap1 = out[0, 1]
heatmap2 = out[0, 2]
heatmap3 = out[0, 3]
plt.imsave("heatmap0.png", heatmap0)
plt.imsave("heatmap1.png", heatmap1)
plt.imsave("heatmap2.png", heatmap2)
plt.imsave("heatmap3.png", heatmap3)
def calc_masked_image(image_filename, object_name, mask_threshold=0.5):
"""
:param image_filename:
:param object_name:
:param mask_threshold:
:return: masked_image, mask, fig, axs
"""
from heatmap.imagenet1000_clsid_to_human import get_imagenet_classes_from_names
class_ids = get_imagenet_classes_from_names()
# model
class_id = class_ids[object_name]
model = ResNet50()
new_model = to_heatmap(model)
# calc
original_image = imread(image_filename)
image = image_proc.load_img(image_filename,
target_size=(IMAGE_HEIGHT, IMAGE_WIDTH))
heatmap = calculate_heatmap(original_image, new_model, image, class_id,
preprocess_input)
#
masked_image, mask = analyse_heatmap(threshold, original_image, heatmap)
fig, axs = display_graphical_results(original_image, heatmap, masked_image)
return masked_image, mask, fig, axs
def helper_test(model):
img_path = "../examples/dog.jpg"
new_model = to_heatmap(model)
# Loading the image
img = image.load_img(img_path, target_size=(800, 800))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
out = new_model.predict(x)
s = "n02084071" # Imagenet code for "dog"
ids = synset_to_dfs_ids(s)
heatmap = out[0]
if K.image_data_format() == 'channels_first':
heatmap = heatmap[ids]
heatmap = np.sum(heatmap, axis=0)
else:
heatmap = heatmap[:, :, ids]
heatmap = np.sum(heatmap, axis=2)
print(heatmap.shape)
assert heatmap.shape[0] == heatmap.shape[1]
K.clear_session()
# If you have more than 8GB of RAM, you can try to increase it.
img = image.load_img(img_path, target_size=(800, 1280))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
if preprocessing is not None:
x = preprocess_input(x)
out = new_model.predict(x)
heatmap = out[0] # Removing batch axis.
if K.image_data_format() == 'channels_first':
heatmap = heatmap[ids]
if heatmap.ndim == 3:
heatmap = np.sum(heatmap, axis=0)
else:
heatmap = heatmap[:, :, ids]
if heatmap.ndim == 3:
heatmap = np.sum(heatmap, axis=2)
plt.imshow(heatmap, interpolation="none")
plt.show()
model = VGG16()
new_model = to_heatmap(model)
s = "n02084071" # Imagenet code for "dog"
ids = synset_to_dfs_ids(s)
display_heatmap(new_model, "./dog.jpg", ids, preprocess_input)
if K.image_data_format() == 'channels_first':
heatmap = heatmap[ids]
if heatmap.ndim == 3:
heatmap = np.sum(heatmap, axis=0)
else:
heatmap = heatmap[:, :, ids]
if heatmap.ndim == 3:
heatmap = np.sum(heatmap, axis=2)
plt.imshow(heatmap, interpolation="none")
plt.show()
model = VGG16()
new_model = to_heatmap(model)
s = "n02084071" # Imagenet code for "dog"
ids = synset_to_dfs_ids(s)
display_heatmap(new_model, "./dog.jpg", ids, preprocess_input)
# Also testing with a sequential model
sequential_model = Sequential()
sequential_model.add(
Conv2D(64, (3, 3),
activation='relu',
padding='same',
input_shape=(224, 224, 3)))
sequential_model.add(Conv2D(64, (3, 3), activation='relu', padding='same'))
sequential_model.add(MaxPooling2D((2, 2), strides=(2, 2)))