def print_funked_scenes(data):
subsampling = data[data["samples"] != data["samples_reference"]]
# scene_names = [ should_accept.get_scene_name( row("image").decode('UTF-8') ) for row in data ]
for scene in should_accept.ok_thresholds:
filtered_ok = subsampling[(
subsampling["samples"] == should_accept.ok_thresholds[scene]) |
(subsampling["samples_reference"] ==
should_accept.ok_thresholds[scene])]
filtered_ok_not_ok = filtered_ok[
(filtered_ok["samples"] == should_accept.not_ok_thresholds[scene])
| (filtered_ok["samples_reference"] ==
should_accept.not_ok_thresholds[scene])]
if scene == "atom":
for row in filtered_ok_not_ok:
if should_accept.get_scene_name(
row["image"].decode('UTF-8')) == scene:
# print ( str( row["crop_x"] ) + ", " + str( row["crop_y"] ) )
print(row)
rows = sum(1 for row in filtered_ok_not_ok
if should_accept.get_scene_name(row["image"].decode(
'UTF-8')) == scene)
print("Scene: " + scene + ", rows: " + str(rows))
def split_by_scenes(data, scenes):
if "scene" in data.dtype.names:
test_mask = data["scene"] == "not_existing_scene_to_create_empty_mask"
for scene in scenes:
test_mask = test_mask | (data["scene"] == scene)
train_mask = ~test_mask
test_set = data[test_mask]
train_set = data[train_mask]
return train_set, test_set
else:
test_set = [
row for row in data
if accept.get_scene_name(row["image"].decode('UTF-8')) in scenes
]
train_set = [
row for row in data if accept.get_scene_name(row["image"].decode(
'UTF-8')) not in scenes
]
return train_set, test_set
def main_loop(config):
global screen
global esc_pressed
global scene_names
data = loading.load_dataset(config.dataset)
full_images = select_rows(data, config)
scene_names = [
should_accept.get_scene_name(row["image"].decode('UTF-8'))
for row in data
]
load_next_row(data, full_images)
screen = numpy.zeros(image.shape, numpy.uint8)
cv2.namedWindow('Crops labeling')
cv2.setMouseCallback('Crops labeling', mouse_select)
while (1):
screen = apply_mask(image, selected_crops)
draw_grid_lines(screen)
cv2.imshow('Crops labeling', screen)
key = cv2.waitKey(20)
if key == ord('m'):
show_hide_mask()
elif key == ord('d'):
update_crops_selection(data, config)
load_next_row(data, full_images)
elif key == ord('a'):
update_crops_selection(data, config)
load_previous_row(data, full_images)
elif key == ord("h"):
print_help()
elif key == ord("l"):
show_grid_lines()
elif key == ord("\r"):
update_crops_selection(data, config)
data_filtering.save_binary(data, config.dataset)
elif key == ord("y"):
if esc_pressed:
update_crops_selection(data, config)
data_filtering.save_binary(data, config.dataset)
break
else:
pass
elif key == ord("n"):
if esc_pressed:
break
else:
pass
elif key == 27:
print("Do you want to save your dataset labeling? (y/n)")
esc_pressed = True
cv2.destroyAllWindows()
def update_crops_selection(data, config):
if config.filter_threshold:
indicies = compute_row_and_all_crops_idx(data)
else:
indicies = compute_row_and_same_samples_crops_idx(data)
scene = should_accept.get_scene_name(current_row["image"].decode('UTF-8'))
for idx in indicies:
row = data[idx]
if row["is_cropped"]:
tile_x = row["crop_x"]
tile_y = row["crop_y"]
label = selected_crops[tile_x][tile_y]
if label == FalseLabel:
data[idx]["label"] = get_label_string(
should_accept.tell_from_samples(
row["image"].decode('UTF-8'),
row["reference_image"].decode('UTF-8')).value)
elif label == TrueLabel:
threshold = should_accept.not_ok_thresholds[scene]
if row["samples"] >= threshold and row[
"samples_reference"] >= threshold:
data[idx]["label"] = b"TRUE"
else:
data[idx]["label"] = b"IGNORE"
else:
pass
def compute_row_and_all_crops_idx(data):
filtered_mask = data["samples"] != data["samples_reference"]
scene = should_accept.get_scene_name(current_row["image"].decode('UTF-8'))
scene_mask = [scene_name == scene for scene_name in scene_names]
return numpy.where(filtered_mask & scene_mask)[0]
def data_sceneing(data):
print("Sceneing data")
scene_names = [
should_accept.get_scene_name(row["image"].decode('UTF-8'))
for row in data
]
data["scene"] = scene_names
return data
def save_all_crops(data, compared_dir, reference_dir):
data = data[data["is_cropped"] == True]
num_crops = 10
for i, scene in enumerate(should_accept.ok_thresholds):
print("Scene number: " + str(i) + ", name: " + scene)
filtered = data[data["samples"] ==
should_accept.not_ok_thresholds[scene]]
filtered = filtered[filtered["samples_reference"] ==
should_accept.ok_thresholds[scene]]
for row in filtered:
if should_accept.get_scene_name(
row["image"].decode('UTF-8')) == scene:
reference_file = os.path.normpath(
row["reference_image"].decode('UTF-8'))
compared_file = os.path.normpath(row["image"].decode('UTF-8'))
reference_file = reference_file.replace(
"D:\\GolemData", "e:\\golem")
compared_file = compared_file.replace("D:\\GolemData",
"e:\\golem")
print("Processing files: reference [" + reference_file +
"] and compared [" + compared_file + "].")
(reference_image,
compared_image) = load_images(reference_file, compared_file)
(cropped_image,
cropped_reference) = extract_features.crop_image(
row["crop_x"], row["crop_y"], num_crops, compared_image,
reference_image)
cropped_image.save(
os.path.join(compared_dir,
gen_file_name(compared_file, row, False)),
"PNG")
cropped_reference.save(
os.path.join(reference_dir,
gen_file_name(reference_file, row, False)),
"PNG")
def analyze_wavelets(data):
mask = [
should_accept.get_scene_name(
row["image"].decode('UTF-8')) == "glass_material" for row in data
]
filtered = data[mask]
filtered = filtered[filtered["is_cropped"] == True]
filtered = filtered[filtered["crop_x"] >= 1]
filtered = filtered[filtered["crop_x"] <= 1]
filtered = filtered[filtered["crop_y"] >= 4]
filtered = filtered[filtered["crop_y"] <= 4]
filtered = filtered[filtered["samples"] != filtered["samples_reference"]]
filtered = filtered[filtered["samples_reference"] == 8325]
plot_data("samples", "wavelet_high", filtered)
plot_data("samples", "wavelet_low", filtered)
plot_data("samples", "wavelet_mid", filtered)
matplotlib.pyplot.show()
def select_scenes_with_threshold(data):
print("Selecting scenes on subsampling threshold.")
filtered = data[data["samples"] != data["samples_reference"]]
# Create mask with all values False
mask = filtered["samples"] == 0
scene_names_filtered = [
should_accept.get_scene_name(row["image"].decode('UTF-8'))
for row in filtered
]
for i, scene in enumerate(should_accept.ok_thresholds):
print("Finding images on threshold for scene: " + scene)
compared_ok_mask = (filtered["samples"]
== should_accept.ok_thresholds[scene]) & (
filtered["samples_reference"]
== should_accept.not_ok_thresholds[scene])
reference_ok_mask = (filtered["samples_reference"]
== should_accept.ok_thresholds[scene]) & (
filtered["samples"]
== should_accept.not_ok_thresholds[scene])
scene_mask = [
scene_name == scene for scene_name in scene_names_filtered
]
threshold_mask = (compared_ok_mask | reference_ok_mask)
scene_threshold_mask = (threshold_mask & scene_mask)
mask = mask | scene_threshold_mask
return filtered[mask]