我創建了一個字符串的 Tensorflow 數據集（其中每個字符串都是 dicom 圖像的路徑），我想將預處理函數映射到數據集。預處理函數應該使用pydicom包從 dicom 文件加載像素數組。但是當我嘗試映射一個函數時，雖然我得到一個屬性錯誤。如何使用如下函數從張量中讀取字符串值？我正在使用 Tensorflow 2.0.0 和 pydicom 1.3.0。AttributeError: in converted code: <ipython-input-12-eff65198c202>:12 load_and_preprocess_image * dicom_data = pydicom.dcmread(img_path) /opt/conda/lib/python3.6/site-packages/pydicom/filereader.py:849 dcmread * dataset = read_partial(fp, stop_when, defer_size=defer_size, /opt/conda/lib/python3.6/site-packages/pydicom/filereader.py:651 read_partial * preamble = read_preamble(fileobj, force) /opt/conda/lib/python3.6/site-packages/pydicom/filereader.py:589 read_preamble * preamble = fp.read(128) AttributeError: 'Tensor' object has no attribute 'read'這是我創建數據集并將預處理函數映射到它的代碼。def load_and_preprocess_image(img_path): """ Load image, resize, and normalize the image""" dicom_data = pydicom.dcmread(img_path)) image = tf.convert_to_tensor(dicom_data.pixel_array, dtype=tf.float32) return image# Create dataset (list of strings that lead to dicom paths)image_train_ds = tf.data.Dataset.from_tensor_slices(dicom_files_list)# Map a preprocessing function to list of dicom pathsimage_train_ds = image_train_ds.map(load_and_preprocess_image)
2 回答
慕尼黑8549860
TA貢獻1818條經驗 獲得超11個贊
有一個 Tensorflow-io 包現在允許這樣做。它相當簡單。
安裝包：
pip install -q tensorflow-io
import tensorflow_io as tfio
def load_and_preprocess_image(img_path):
_bytes = tf.io.read_file(img_path)
dicom_data = tfio.image.decode_dicom_image(_bytes, dtype=tf.float32)
return dicom_data
dicom_files_list = ['path/to/dicom']
# Create dataset (list of strings that lead to dicom paths)
image_train_ds = tf.data.Dataset.from_tensor_slices(dicom_files_list)
image_train_ds = image_train_ds.map(load_and_preprocess_image)
弒天下
TA貢獻1818條經驗 獲得超8個贊
在 中pydicom.dcmread(img_path)，img_path是 tf.string 張量。我不認為pydicom支持讀取張量對象。
我找到了一種解決方法，它是在 tensorflow 中提供 DICOM 操作的 gradient_decode_dicom 。以下代碼改編自此 colab，并在 tf2.0 上進行了測試。
def load_and_preprocess_image(img_path):
_bytes = tf.io.read_file(img_path)
dicom_data = decode_dicom_image(_bytes, dtype=tf.float32)
return dicom_data
dicom_files_list = ['path/to/dicom']
# Create dataset (list of strings that lead to dicom paths)
image_train_ds = tf.data.Dataset.from_tensor_slices(dicom_files_list)
image_train_ds = image_train_ds.map(load_and_preprocess_image)
添加回答
舉報
0/150
提交
取消