def main(image, image_name, rows, columns, filter_list, csv_output,
yaml_out=None, save_cell_images=False, filter_image_name=None):
cell_w, cell_h = dims.celldims(image.size, rows, columns)
filter_name = "-".join(map(base_name_no_ext, args.filters))
filtered_image = apply_filters(image, filter_list)
if filter_image_name:
filtered_image.save(filter_image_name)
stats_data = []
vector_data = []
stat_functions = dict(mean=chi_mean, stddev=np.std, peaks=count_max_peaks)
for i in range(rows):
stats_data.append(list())
vector_data.append(list())
for j in range(columns):
stats_data[i].append(dict())
bb = bounding_box(j * cell_w, i * cell_h, cell_w, cell_h)
cell = filtered_image.crop(bb)
if save_cell_images:
cell_name = "{:s}-{:s}_y{:02d}_x{:02d}.bmp".format(image_name, filter_name, i, j)
cell.save(cell_name)
vectors = get_vectors(cell)
for key in vectors:
# union of two dict to add new keys
stats_data[i][j].update(dict_map(vectors[key], stat_functions, key + "_"))
vector_data[i].append(vectors)
if yaml_out: # yaml lib doesn't like numpy data types.
vectors['x'] = list(map(float, vectors['x']))
vectors['y'] = list(map(float, vectors['x']))
if yaml_out:
yaml_out.write(yaml_dump(vector_data))
stat_names = stats_data[0][0].keys()
header_names = [filter_name + "-" + stat_name for stat_name in stat_names]
stats_writer = csv.writer(csv_output)
stats_writer.writerow(['i', 'j', 'x', 'y'] + header_names)
for i in range(rows):
for j in range(columns):
row = [i, j, float(j) / columns, float(i) / rows]
for k in stat_names:
row.append(stats_data[i][j][k])
stats_writer.writerow(row)
if __name__ == "__main__":
args = parse_args()
if args.gold:
meta_data = yaml.load(open(args.gold))
for name, data in meta_data.items():
meta_data[name]['image'] = Image.open(data['path'])
meta_data[name]['template'] = Image.open(data['template_path'])
else:
meta_data = dict()
grid = Image.open(args.map)
classes = [r for r in csv.reader(open(args.intended))]
rows = len(classes)
cols = len(classes[0])
width, height = dims.celldims(grid.size, rows, cols)
# collect all cells by class name
img_by_class = dict()
for j in range(rows):
for i in range(cols):
name = classes[j][i]
cell = grid.crop(bounding_box(i * width, j * height, width, height))
if name in img_by_class:
img_by_class[name].append(cell)
else:
img_by_class[name] = [cell]
# align all the images to the first
if not args.noalign:
for name, imgs in img_by_class.items():
parser.add_argument('-split', type=int)
return parser.parse_args()
def gold_classify(gold_images, cell_image, align=True):
return [(compare_images(gold, cell_image, align=False), className) for className, gold in gold_images.items()]
def bounding_box(x, y, w, h, inset=7):
return (x + inset, y + inset, x + w - 2 * inset, y + h - 2 * inset)
if __name__ == "__main__":
args = parse_args()
grid = Image.open(args.image)
cell_w, cell_h = dims.celldims(grid.size, args.rows, args.cols)
answer = []
gold = yaml.load(open(args.gold))
for name, mean_img in gold.items():
if 'path' in mean_img:
gold[name] = Image.open(mean_img['path'])
else:
gold[name] = Image.open(mean_img)
if args.split:
gold[name] = as_black_and_white(gold[name], args.split, contrast=True)
class_errs = [[gold_classify(gold,
grid.crop(bounding_box(
j * cell_w, i * cell_h, cell_w, cell_h)))
for j in range(args.cols)] for i in range(args.rows)]
