IndexError: index 28 is out of bounds for axis 0 with size 4 when run crop_mouth_from_video.py

[SHAIIM04]

Thanks to release your code to public, I use it own my dataset first a extract landmark from https://github.com/mpc001/Visual_Speech_Recognition_for_Multiple_Languages/blob/master/crop_mouth.py this work then fed them with data dir to extract mouth region by using preprocessing/crop_mouth_from_video.py but there is error apper , I don't know if landmark it is a problem or not .

Traceback (most recent call last):
File "crop_mouth_from_video.py", line 47, in
sequence = dataloader.load_data(
File "/Users/shaimaa/Downloads/LIP_Reading/Code/audio-visual-dataset-master/Lipreading_using_Temporal_Convolutional_Networks/preprocessing/dataloader.py", line 195, in load_data
return self.load_video(data_filename, landmarks_filename=landmarks_filename)
File "/Users/shaimaa/Downloads/LIP_Reading/Code/audio-visual-dataset-master/Lipreading_using_Temporal_Convolutional_Networks/preprocessing/dataloader.py", line 179, in load_video
sequence = self.preprocess(
File "/Users/shaimaa/Downloads/LIP_Reading/Code/audio-visual-dataset-master/Lipreading_using_Temporal_Convolutional_Networks/preprocessing/dataloader.py", line 65, in preprocess
sequence, transformed_frame, transformed_landmarks = self.crop_patch(video_pathname, preprocessed_landmarks)
File "/Users/shaimaa/Downloads/LIP_Reading/Code/audio-visual-dataset-master/Lipreading_using_Temporal_Convolutional_Networks/preprocessing/dataloader.py", line 117, in crop_patch
transformed_frame, transformed_landmarks = self.affine_transform(frame, smoothed_landmarks, self._reference, grayscale=self._convert_gray)
File "/Users/shaimaa/Downloads/LIP_Reading/Code/audio-visual-dataset-master/Lipreading_using_Temporal_Convolutional_Networks/preprocessing/dataloader.py", line 155, in affine_transform
transform = cv2.estimateAffinePartial2D(np.vstack([landmarks[x] for x in stable_points]),
File "/Users/shaimaa/Downloads/LIP_Reading/Code/audio-visual-dataset-master/Lipreading_using_Temporal_Convolutional_Networks/preprocessing/dataloader.py", line 155, in
transform = cv2.estimateAffinePartial2D(np.vstack([landmarks[x] for x in stable_points]),
IndexError: index 28 is out of bounds for axis 0 with size 4