timestamp = time.strftime('_%m%d%H%M%S', time.localtime())
output_dir = './test_result/test_result' + timestamp
check_output_dir(output_dir)
test_data = './test'
list_dir = os.listdir(test_data)
for target in list_dir:
# read one image
imagePath = os.path.join(test_data, target)
print(imagePath)
image = cv2.imread(imagePath)
success, local = face_detection(image)
if success:
local = box_reduce(local, 30)
img_32 = image_pre_processing(image, local)
result, guess = img_test(model, img_32) # predict
pd_result_name = map_characters[result]
print('index:', result)
print('name:', pd_result_name)
cv2.rectangle(image, (local[0], local[1]), (local[2], local[3]),
(0, 255, 0), 4, cv2.LINE_AA)
cv2.putText(image, pd_result_name, (local[0], local[1] - 8),
cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 255), 2,
cv2.LINE_AA)
timestamp = str(int(round(time.time() * 1000)))
cv2.imwrite(output_dir + '/' + str(guess) + '_' + str(target),
# write predict result name
pd_result_name = ''
count_list = [0] * len(anchor_name) ###
while videoIn.isOpened():
# read video from camera
ret, outframe = videoIn.read()
if (ret):
# keyboard input value
key = cv2.waitKey(1) & 0xFF
success, local = face_detection(outframe)
if success:
local = box_reduce(local)
img_32 = image_pre_processing(outframe, local)
#--------------- predict ----------------
img_32_np = np.array(img_32, dtype='float32')
img_32_np = img_32_np[np.newaxis, np.newaxis, :, :] / 255.0
img_feature = fix_model.predict(img_32_np)
img_feature = img_feature[np.newaxis, :, :, :, :]
fix_feature_repeat = np.repeat(fix_feature,
img_feature.shape[1],
axis=0)
img_feature_repeat = np.repeat(img_feature,
fix_feature.shape[1],
