def main():
filename = []
for i in range(0, 4):
tmp = "../sample/camera_photo/0317_test/" + str(i + 1) + ".jpg"
filename.append(tmp)
## 5 classes model
classes = [
'musical_symbol_bass_clef', 'musical_symbol_half_note',
'musical_symbol_quarter_note', 'musical_symbol_quarter_rest',
'musical_symbol_g_clef'
]
model_number = "0401_9"
model_dir = "../Model/" + model_number + "/"
model_name = "model_5_classes_" + model_number + ".meta"
true_label_file = []
for i in range(0, 4):
tmp = "../sample/true_label/0317_test/" + str(
i + 1) + "_true_label_5_classes.csv"
true_label_file.append(tmp)
t_acc_matrix = init_Conf_mat.InitConfMat(classes)
#for i in range(0, 1):
for i in range(0, 4):
y_position, y_position_bottom = five.five_lines(filename[i])
symbol_Info = x_scan.x_cut(y_position, y_position_bottom, filename[i])
allInfo_symbol = predict.predict_symbol(symbol_Info, filename[i],
classes, model_dir, model_name)
allInfo_symbol, t_acc_matrix = accuracy.acc(allInfo_symbol,
true_label_file[i],
classes, filename[i],
t_acc_matrix, model_name)
OutputResult.Output(allInfo_symbol, classes, filename[i], model_name)
#pitch.Identify(allInfo_symbol, filename[i], y_position, y_position_bottom);
summary.Summary(t_acc_matrix, model_name)
def one_graph(prob_missingness):
p = prob_missingness
without_imp = np.zeros(5)
grand_mean = np.zeros(5)
conditional_mean = np.zeros(5)
closest = np.zeros(5)
regression = np.zeros(5)
without_removing = np.zeros(5)
multiple_closest = np.zeros(5)
multiple_regression = np.zeros(5)
training_set_size = ['50', '100', '250', '500', '1000']
dini = (pandas.read_csv('data_banknote_authentication.csv').to_numpy())
np.random.shuffle(dini)
X1i = dini[0:1000, 0:4]
X2i = dini[1000:, 0:4]
Y1i = dini[0:1000, 4]
Y2i = dini[1000:, 4]
for i in range(nb_tests):
df = (pandas.read_csv('data_banknote_authentication.csv').to_numpy())
np.random.shuffle(df)
for i in range(1371):
for j in range(4):
if ((np.random.uniform()) < p * 2) and (df[i, 4] == 1.0):
df[i, j] = 0
fullX1 = df[0:1000, 0:4]
fullY1 = df[0:1000, 4]
X2 = df[1000:, 0:4]
Y2 = df[1000:, 4]
for j in range(5):
without_removing[j] += acc.acc(X1i[0:used_trs[j]][:],
Y1i[0:used_trs[j]], X2i, Y2i,
'no_imputation')
without_imp[j] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2,
'no_imputation')
grand_mean[j] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2,
'grand_mean')
conditional_mean[j] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2,
'conditional_mean')
closest[j] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2, 'closest')
regression[j] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2,
'regression')
multiple_closest[j] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2,
'multiple_closest')
multiple_regression[j] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2,
'multiple_regression')
without_removing /= nb_tests
without_imp /= nb_tests
grand_mean /= nb_tests
conditional_mean /= nb_tests
closest /= nb_tests
regression /= nb_tests
multiple_closest /= nb_tests
multiple_regression /= nb_tests
plt.figure(figsize=(10, 5))
plt.plot(training_set_size, without_removing, color='purple', linewidth=2)
plt.plot(training_set_size, without_imp, color='green', linewidth=2)
plt.plot(training_set_size, grand_mean, color='blue', linewidth=2)
plt.plot(training_set_size, conditional_mean, color='orange', linewidth=2)
plt.plot(training_set_size, closest, color='red', linewidth=2)
plt.plot(training_set_size, regression, color='black', linewidth=2)
plt.plot(training_set_size, multiple_closest, color='pink', linewidth=2)
plt.plot(training_set_size, multiple_regression, color='grey', linewidth=2)
mylabels = [
'without_removing', 'No imputation', 'Grand Mean', 'Conditional Mean',
'Closest neighbour', 'Regression', 'multiple_closest',
'multiple_regression'
]
#,'multiple_closest','multiple_regression'
plt.title(' Real_life_data MNAR ' + ' Prob_missingness: ' + str(p))
plt.legend(labels=mylabels)
plt.ylabel('Accuracy', fontsize=10)
plt.xlabel('Training size', fontsize=10)
plt.show()
def one_graph(ncov, dim, type_missingness, prob_missingness):
p = prob_missingness
without_imp = np.zeros(6)
grand_mean = np.zeros(6)
conditional_mean = np.zeros(6)
closest = np.zeros(6)
regression = np.zeros(6)
without_removing = np.zeros(6)
#multiple_closest=np.zeros(6)
#multiple_regression=np.zeros(6)
training_set_size = ['50', '100', '250', '500', '1000', '2000']
for i in range(nb_tests):
data = dt.full_gen(ncov, dim, 2000, 1000, type_missingness,
prob_missingness)
fullX1 = data[0]
fullY1 = data[1]
X2 = data[2]
Y2 = data[3]
X1i = data[4]
X2i = data[5]
for j in range(6):
without_removing[j] += acc.acc(X1i[0:used_trs[j]][:],
fullY1[0:used_trs[j]], X2i, Y2,
'no_imputation')
without_imp[j] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2,
'no_imputation')
grand_mean[j] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2,
'grand_mean')
conditional_mean[j] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2,
'conditional_mean')
closest[j] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2, 'closest')
regression[j] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2,
'regression')
#multiple_closest[j]+=acc.acc(fullX1[0:used_trs[j],:],fullY1[0:used_trs[j]],X2,Y2,'multiple_closest')
#multiple_regression[j]+=acc.acc(fullX1[0:used_trs[j],:],fullY1[0:used_trs[j]],X2,Y2,'multiple_regression')
without_removing /= nb_tests
without_imp /= nb_tests
grand_mean /= nb_tests
conditional_mean /= nb_tests
closest /= nb_tests
regression /= nb_tests
#multiple_closest/=nb_tests
#multiple_regression/=nb_tests
plt.figure(figsize=(10, 5))
plt.plot(training_set_size, without_removing, color='purple', linewidth=2)
plt.plot(training_set_size, without_imp, color='green', linewidth=2)
plt.plot(training_set_size, grand_mean, color='blue', linewidth=2)
plt.plot(training_set_size, conditional_mean, color='orange', linewidth=2)
plt.plot(training_set_size, closest, color='red', linewidth=2)
plt.plot(training_set_size, regression, color='black', linewidth=2)
#plt.plot(training_set_size,multiple_closest, color = 'pink', linewidth = 2)
#plt.plot(training_set_size,multiple_regression, color = 'grey', linewidth = 2)
mylabels = [
'without_removing', 'No imputation', 'Grand Mean', 'Conditional Mean',
'Closest neighbour', 'Regression'
]
#,'multiple_closest','multiple_regression'
plt.title('Covariance: ' + ncov + ' Dim: ' + str(dim) +
' Type missingness: ' + type_missingness +
' Prob_missingness: ' + str(p))
plt.legend(labels=mylabels)
plt.ylabel('Accuracy', fontsize=10)
plt.xlabel('Training size', fontsize=10)
plt.axis([0, 6, 0.7, 0.95])
plt.show()
config['link_conf_thr'] = 0.8
config['min_area'] = 300
config['min_height'] = 10
i = 0
total_time1 = time.time()
for impath in impaths:
#impath = '/home/blin/Downloads/text_detection/test/1-123152001-OCR-LF-C01.jpg'
#impath = '/home/blin/Downloads/text_detection/test/1-142434001-OCR-AH-A01.jpg'
imname = os.path.basename(impath)
im = cv2.imread(impath)
print(impath)
t1 = time.time()
bboxs = detection(im, sess_d, input_x, segm_logits, link_logits, config)
t2 = time.time()
print('detection_time: ', (t2 - t1), 'result', bboxs)
#bboxs = ['792, 364, 792, 298, 923, 298, 923, 364\n', '972, 375, 972, 303, 1271, 303, 1271, 375\n', '972, 455, 972, 389, 1109, 389, 1109, 455\n']
predicted = recognition(im, sess_r_h, sess_r_v, bboxs, (240, 32),
images_ph_h, images_ph_v, model_out_h, model_out_v,
decoded_h, decoded_v)
#predicted = recognition(im, sess_r_h , bboxs, (240, 32), images_ph_h, model_out_h, decoded_h)
t3 = time.time()
print('recognition_time: ', (t3 - t2), 'result', predicted)
i += 1
print(i)
line = imname + ' ' + predicted + '\n'
res_txt.write(line)
res_txt.close()
total_time2 = time.time()
print('total_time: ', (total_time2 - total_time1))
acc('/home/blin/containernumber_result.txt')
def one_graph(ncov, dim, type_missingness, prob_missingness):
p = prob_missingness
without_imp = np.zeros((6, nb_tests))
grand_mean = np.zeros((6, nb_tests))
conditional_mean = np.zeros((6, nb_tests))
closest = np.zeros((6, nb_tests))
regression = np.zeros((6, nb_tests))
without_removing = np.zeros((6, nb_tests))
training_set_size = ['50', '100', '250', '500', '1000', '2000']
for i in range(nb_tests):
data = dt.full_gen(ncov, dim, 2000, 1000, type_missingness,
prob_missingness)
fullX1 = data[0]
fullY1 = data[1]
X2 = data[2]
Y2 = data[3]
X1i = data[4]
X2i = data[5]
for j in range(6):
without_removing[j][i] += acc.acc(X1i[0:used_trs[j]][:],
fullY1[0:used_trs[j]], X2i, Y2,
'no_imputation')
without_imp[j][i] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2,
'no_imputation')
grand_mean[j][i] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2,
'grand_mean')
conditional_mean[j][i] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2,
'conditional_mean')
closest[j][i] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2, 'closest')
regression[j][i] += acc.acc(fullX1[0:used_trs[j], :],
fullY1[0:used_trs[j]], X2, Y2,
'regression')
v_without_removing = np.var(without_removing, axis=1)
v_without_imp = np.var(without_imp, axis=1)
v_grand_mean = np.var(grand_mean, axis=1)
v_conditional_mean = np.var(conditional_mean, axis=1)
v_closest = np.var(closest, axis=1)
v_regression = np.var(regression, axis=1)
plt.figure(figsize=(10, 5))
plt.plot(training_set_size,
v_without_removing,
color='purple',
linewidth=2)
plt.plot(training_set_size, v_without_imp, color='green', linewidth=2)
plt.plot(training_set_size, v_grand_mean, color='blue', linewidth=2)
plt.plot(training_set_size,
v_conditional_mean,
color='orange',
linewidth=2)
plt.plot(training_set_size, v_closest, color='red', linewidth=2)
plt.plot(training_set_size, v_regression, color='black', linewidth=2)
mylabels = [
'without_removing', 'No imputation', 'Grand Mean', 'Conditional Mean',
'Closest neighbour', 'Regression'
]
plt.title('Accuracy Variance' + 'Covariance: ' + ncov + ' Dim: ' +
str(dim) + ' Type missingness: ' + type_missingness +
' Prob_missingness: ' + str(p))
plt.legend(labels=mylabels)
plt.ylabel('Accuracy', fontsize=10)
plt.xlabel('Training size', fontsize=10)
plt.savefig('Variance ' + 'Covariance: ' + ncov + ' Dim: ' + str(dim) +
' Type missingness: ' + type_missingness +
' Prob_missingness: ' + str(p) + '.jpg')
plt.show()