Keras 无法通过“flow_from_dataframe”找到图像

Question

I'm trying run a ImageDataGenerator using the code below but I keep running into the error;

"Found 0 validated image filenames."

The most obvious point would be an error in the paths (directory and x_col) but I'm able to read the files using imageio/pydicom with the corresponding paths from the code below. Another point could be the fact that I am trying to read Dicom (DCM) images but I cannot find a source to substantiate this probability.

I have been trying solutions for the last hour and am starting to get slightly frustrated. What am I missing?

train_data = train_data_generator.flow_from_dataframe(
    dataframe = data_input,
    directory = input_dir_im,
    x_col = 'path',
    y_col = class_list,
    target_size = input_tensor,
    color_mode = 'grayscale',
    batch_size = batch_size,
    shuffle = False,
    class_mode = 'raw',
    )  

Answer 1

就像@Dr.Snoopy 指出的那样，Keras 仅支持标准图像文件格式。