def generate_line():
line_data = _generate_line("line")
line_data['type'] = "line"
line_data['qa_pairs'] = generate_line_plot_questions(
combine_source_and_rendered_data(line_data), color_map=color_map)
visuals = _generate_visuals_for_line_plot(line_data['data'])
# Add variation for line styles
solid_only = True if np.random.random(
) <= data_config['line']['solid_pr'] else False
if solid_only:
line_styles = ["solid"] * len(line_data['data'])
else:
reference_styles = ["solid", "dashed", "dotted", "dotdash", "dashdot"]
permuted_styles = list(np.random.permutation(reference_styles))
line_styles = permuted_styles[:]
while len(line_styles) < len(line_data['data']):
line_styles += permuted_styles
line_styles = line_styles[:len(line_data['data'])]
visuals['line_styles'] = line_styles
line_data['visuals'] = visuals
return line_data
def generate_dot_line():
line_data = _generate_line("dot_line")
line_data['type'] = "dot_line"
line_data['qa_pairs'] = generate_line_plot_questions(
combine_source_and_rendered_data(line_data), color_map=color_map)
line_data['visuals'] = _generate_visuals_for_line_plot(line_data['data'])
return line_data
def generate_hbar_categorical():
bar_data = _generate_bar_categorical("hbar_categorical")
old_x = bar_data['data'][0]['x']
bar_data['data'][0]['x'] = bar_data['data'][0]['y']
bar_data['data'][0]['y'] = old_x
bar_data['type'] = "hbar_categorical"
bar_data['qa_pairs'] = generate_bar_graph_questions(
combine_source_and_rendered_data(bar_data), color_map=color_map)
return bar_data
def generate_figures(source_data_json,
destination_directory,
add_bboxes=False,
supplied_webdriver=None):
# Setup dest dirs
qa_json_dir = os.path.join(destination_directory, "json_qa")
annotations_json_dir = os.path.join(destination_directory,
"json_annotations")
html_dir = destination_directory
png_dir = os.path.join(destination_directory, "png")
dirs = [destination_directory, qa_json_dir, annotations_json_dir, png_dir]
if add_bboxes:
bbox_img_dir = os.path.join(destination_directory, "bbox_png")
dirs.append(bbox_img_dir)
for dirpath in dirs:
if not os.path.exists(dirpath):
os.mkdir(dirpath)
# Create web driver
if supplied_webdriver:
webdriver = supplied_webdriver
else:
webdriver = seldriver.PhantomJS()
# Read in the synthetic data
with open(source_data_json, 'r') as f:
source_data_json = json.load(f)
for fig_id, source in tqdm(iter(enumerate(source_data_json['data'])),
total=len(source_data_json['data']),
desc="Plotting figures"):
point_sets = source['data']
fig = None
fig_type = source['type']
if fig_type == 'vbar_categorical':
fig = VBarGraphCategorical(point_sets[0], source['visuals'])
elif fig_type == 'hbar_categorical':
fig = HBarGraphCategorical(point_sets[0], source['visuals'])
elif fig_type == 'line':
fig = LinePlot(point_sets, source['visuals'])
elif fig_type == 'dot_line':
fig = DotLinePlot(point_sets, source['visuals'])
elif fig_type == 'pie':
fig = Pie(point_sets[0], source['visuals'])
if not fig:
continue
html_file = os.path.join(html_dir, "%d_%s.html" % (fig_id, fig_type))
png_file = os.path.join(png_dir, "%d_%s.png" % (fig_id, fig_type))
# Export to HTML, PNG, and get rendered data
rendered_data = export_png_and_data(fig.figure, png_file, html_file,
webdriver)
all_plot_data = combine_source_and_rendered_data(source, rendered_data)
qa_json_file = os.path.join(qa_json_dir,
"%s_%s.json" % (fig_id, fig_type))
annotations_json_file = os.path.join(
annotations_json_dir,
"%d_%s_annotations.json" % (fig_id, fig_type))
for qa in source['qa_pairs']:
qa['image'] = os.path.basename(png_file)
qa['annotations'] = os.path.basename(annotations_json_file)
with open(qa_json_file, 'w') as f:
json.dump(
{
'qa_pairs':
source['qa_pairs'],
'total_distinct_questions':
source_data_json['total_distinct_questions'],
'total_distinct_colors':
source_data_json['total_distinct_colors']
}, f)
with open(annotations_json_file, 'w') as f:
json.dump(all_plot_data, f)
if add_bboxes:
all_plot_data['image_index'] = fig_id
generate_all_images_with_bboxes_for_plot(all_plot_data,
png_file,
bbox_img_dir,
'red',
load_image=True)
# Cleanup
os.remove(html_file)
# Kill the newly created webdriver
if not supplied_webdriver:
webdriver.service.process.send_signal(signal.SIGTERM)
try:
RemoteWebDriver.quit(webdriver)
except:
pass
def generate_pie():
config = data_config['pie']
s, e = config['n_classes_range']
n_classes = np.random.random_integers(s, e)
widths = np.array([np.random.random() + 0.05 for i in range(n_classes)])
widths_radians = 2 * np.pi * widths / np.sum(widths)
starts = [0]
for i in range(0, n_classes - 1):
starts.append(starts[i] + widths_radians[i])
ends = starts[1:] + [2 * np.pi]
thetas = [
starts[i] + (ends[i] - starts[i]) / 2 for i in range(len(starts))
]
rad = 0.75
x = [rad * np.cos(theta) for theta in thetas]
y = [rad * np.sin(theta) for theta in thetas]
# Get colors and labels
all_color_pairs = []
with open(os.path.normpath(config['color_sources'][0]), 'r') as f:
for w in f.readlines():
name, color = w.split(',')
all_color_pairs.append((name.strip(), color.strip()))
selected_color_pairs = random.sample(all_color_pairs, n_classes)
pie_data = {
'type':
"pie",
'data': [{
'label_x': x,
'label_y': y,
'labels': [cp[0] for cp in selected_color_pairs],
'colors': [cp[1] for cp in selected_color_pairs],
'spans': widths_radians.tolist(),
'starts': starts,
'ends': ends,
}]
}
# Add visuals and legend placement
visuals = _generate_visuals_common()
if visuals['draw_legend']:
# Decide on legend orientation
if n_classes <= common_config[
'legend_horizontal_max_classes'] and np.random.random(
) <= common_config['legend_horizontal_pr']:
visuals['legend_orientation'] = "horizontal"
outside_possibilities = [('below', 'bottom_left'),
('below', 'bottom_center'),
('below', 'bottom_right')]
else:
visuals['legend_orientation'] = "vertical"
outside_possibilities = [('right', 'bottom_right'),
('right', 'center_right'),
('right', 'top_right'),
('left', 'bottom_left'),
('left', 'center_left'),
('left', 'top_left')]
legend_layout_position, legend_position = random.sample(
outside_possibilities, 1)[0]
visuals['legend_position'] = legend_position
visuals['legend_layout_position'] = legend_layout_position
pie_data['visuals'] = visuals
pie_data['qa_pairs'] = generate_pie_chart_questions(
combine_source_and_rendered_data(pie_data), color_map=color_map)
return pie_data
def generate_vbar_categorical():
bar_data = _generate_bar_categorical("vbar_categorical")
bar_data['type'] = "vbar_categorical"
bar_data['qa_pairs'] = generate_bar_graph_questions(
combine_source_and_rendered_data(bar_data), color_map=color_map)
return bar_data