A DICOM reader sequence object using Tensorflow Keras Sequence API

Python_generators--DICOM_reader_seqeunce--04.py

class DicomGenerator(tf.keras.utils.Sequence):
    def __init__(self, dicom_path, batch_size=1, dtype='float32',
                shuffle=False, drop_remainder=False, 
                preserve_batch_size=False, **kwargs):
        self.i = 0
        self.batch_size = batch_size
        self.dtype = dtype
        self.n = len(dicom_path)
        self.dicom_path = dicom_path
        self.drop_remainder = drop_remainder
        self.preserve_batch_size = preserve_batch_size
        self.shuffle = shuffle
        self.__dict__.update(**kwargs)
        self.kwargs = kwargs
  
    def __load__(self, dicom_filename):
        _img = read_dicom(dicom_filename)
        if _img.max() - _img.min() > 0: _img = (_img - _img.min()) / (_img.max() - _img.min())
        _img = np.asarray(_img, dtype=getattr(np, self.dtype))
        return _img

    def __getitem__(self, index):
        if (index + 1) * self.batch_size <= len(self.dicom_path):
            _dicom_path_batch = self.dicom_path[index * self.batch_size:(index + 1) * self.batch_size]
        elif self.drop_remainder:
            raise StopIteration()
        elif self.preserve_batch_size and len(self.dicom_path[index * self.batch_size:len(self.dicom_path)]) !=  self.batch_size:
            _dicom_path_batch = self.dicom_path[len(self.dicom_path)-self.batch_size:len(self.dicom_path)]
        else:
            _dicom_path_batch = self.dicom_path[index * self.batch_size:len(self.dicom_path)]
 
        _img_arr = list(map(lambda _dcm_pth: self.__load__(_dcm_pth), _dicom_path_batch))
        _img_arr = np.array(_img_arr, dtype=getattr(np, self.dtype))
        return _img_arr

    def __iter__(self):
        return self

    def __next__(self):
        if self.i*self.batch_size < len(self.dicom_path):
            _img_arr = self.__getitem__(self.i)
            self.i += 1
        else: raise StopIteration()
        _img_arr = tf.cast(_img_arr, dtype=getattr(tf, self.dtype))
        return _img_arr
    
    def __call__(self):
        self.i = 0
        return self

    def on_epoch_end(self):
        if self.shuffle: random.SystemRandom().shuffle(self.dicom_path)
                
    def __len__(self):
        return self.n // self.batch_size