if feature_type == "HogFeature": feature = HogFeature(mu, sigma) else: feature = HogSizeFeature(mu, sigma) print "classify for label %d with %s"%(label, classifier_file) cnt = 0 failure = 0 if os.path.isdir(src): for filename in os.listdir(src): path = os.path.join(src, filename) img = cv2.imread(path, cv2.CV_LOAD_IMAGE_GRAYSCALE) if img is not None: feature_value = feature.calc(img) label_predict = int(clf.predict(feature_value)) cnt += 1 if label_predict is not label: failure += 1 #print "%d:%s"%(label_predict, path) if SAVE_FAILURES is True: result_path = os.path.join(RESULT_DIR, "%d_%s"%(label_predict,filename)) cv2.imwrite(result_path, img) score = (cnt - failure)* 100/cnt print "score: %d (cnt, failure) = (%d, %d)"%(score, cnt, failure) else: img = cv2.imread(src, cv2.CV_LOAD_IMAGE_GRAYSCALE) if img is not None:
