def eval_dataset(input_dir, output_file, need_latex=True, sketch=False):
recognizer = ShapeRecognizer()
confusion_matrix = {}
for shape in shape_list:
confusion_matrix[shape] = {}
for label in shape_list:
confusion_matrix[shape][label] = 0
error_list = []
for (root, subdirs, filenames) in os.walk(input_dir):
for filename in filenames:
if not filename.endswith('.png'):
continue
print 'Recognizing {0}'.format(filename)
index, label = filename.split('.')[0].split('_')
input_file = os.path.join(root, filename)
try:
recognized_label = recognizer.recognize(input_file)
except:
recognized_label = 'FAIL'
label = label.title()
recognized_label = recognized_label.title()
if label != recognized_label:
feature_str = str(process_image(input_file))
error_list.append((filename, label, recognized_label, feature_str))
confusion_matrix[label][recognized_label] = 1 + confusion_matrix[label].get(recognized_label, 0)
confusion_matrix_percent = copy.deepcopy(confusion_matrix)
for shape in shape_list:
count = sum(confusion_matrix_percent[shape].values())
if count == 0 or sketch:
continue
for key in confusion_matrix_percent[shape].keys():
confusion_matrix_percent[shape][key] /= float(count)
latex_lines = []
with open(output_file + '.csv', 'w') as f:
f.write('Confusion Matrix\n')
f.write(','.join([' '] + shape_list) + '\n')
latex_lines.append(' & '.join([r'\backslashbox{Label}{Recognized}'] + shape_list) + ' \\\\ \\hline\n')
for shape in shape_list:
row = [confusion_matrix_percent[shape][label] for label in shape_list]
line = ','.join([shape] + [('{0:.2f}' if not sketch else '{0}').format(val) for val in row])
latex_lines.append(' & '.join([shape] + [('{0:.2f}' if not sketch else '{0}').format(val) for val in row]) + ' \\\\ \\hline\n')
f.write(line + '\n')
f.write('\n')
f.write('Error List\n')
f.write(','.join(['File', 'Label', 'Recognized', 'Features']) + '\n')
for (filename, label, recognized_label, feature_str) in error_list:
line = ','.join([filename, label, recognized_label, feature_str])
f.write(line + '\n')
if need_latex:
with open(output_file + '_table.tex', 'w') as f:
f.writelines(latex_lines)
import os
import extract_features
import cv2
rootdir = '../data/articles'
for root, dirs, files in os.walk(rootdir, topdown=True):
for file in files:
if file[-3:] == 'jpg':
print 'Processing %s' % os.path.join(root, file)
try:
extract_features.process_image(os.path.join(root, file))
except cv2.error:
pass
import os
import extract_features
import cv2
rootdir = '../data/articles'
for root, dirs, files in os.walk(rootdir, topdown=True):
for file in files:
if file[-3:] == 'jpg':
print 'Processing %s' % os.path.join(root, file)
try:
extract_features.process_image(os.path.join(root, file))
except cv2.error:
pass
def recognize(self, img_file, multi=False):
inputs = process_image(img_file, multi)
result = self.inferencer.evaluate(inputs)
return result['Shape']
