TypeError: If class_mode="categorical", y_col column values must be type string, list or tuple

Question

In keras documentation, it is mentioned that "categorical": 2D numpy array of one-hot encoded labels. Supports multi-label output. but in my case, it does not support.

train_generator_1 = train_data_gen.flow_from_dataframe(annotation,
                                                      directory="data",
                                                      target_size=(img_shape,img_shape),
                                                      x_col="Left",
                                                      y_col=['N','D','G','C','A','H','M','O'],
                                                      class_mode='categorical',
                                                      shuffle=False,
                                                      batch_size=batch_size,
                                                      seed=7)

The picture of labels is attached. Error i am getting is TypeError: If class_mode="categorical", y_col="['N', 'D', 'G', 'C', 'A', 'H', 'M', 'O']" column values must be type string, list or tuple.

Check either y_col does contain any nan value or refer this link to match type of your data github.com/keras-team/keras-preprocessing/blob/… — Welcome_back
– Welcome_back, Commented Feb 9, 2020 at 7:29

Yidong Fu · Accepted Answer · 2020-03-29 22:39:09Z

3

If you have multiple outputs, you'd have to choose class_mode as 'raw'. I found a great tutorial by Vijayabhaskar J here. https://medium.com/@vijayabhaskar96/multi-label-image-classification-tutorial-with-keras-imagedatagenerator-cd541f8eaf24

answered Mar 29, 2020 at 22:39

Yidong Fu

312 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

HanClinto · Accepted Answer · 2022-07-20 16:01:22Z

In my case, I didn't want to switch to raw class_mode, so found a different solution that still uses categorical class_mode.

flow_from_dataframe gets weird when using multiple column names in y_col -- especially when using 'categorical' class_mode. It's expecting the value of the column specified in y_col to be an array -- not for there to be an array of columns. Change it to a single combined column that contains an array of each value, and you'll be golden:

y_cols = ['N','D','G','C','A','H','M','O']
dataframe['classes'] = dataframe[y_cols].apply(lambda x: x.tolist(), axis=1)

train_generator_1 = train_data_gen.flow_from_dataframe(annotation,
                                                      directory="data",
                                                      target_size=(img_shape,img_shape),
                                                      x_col="Left",
                                                      y_col="classes",
                                                      class_mode='categorical',
                                                      shuffle=False,
                                                      batch_size=batch_size,
                                                      seed=7)

Collectives™ on Stack Overflow

TypeError: If class_mode="categorical", y_col column values must be type string, list or tuple

2 Answers 2

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related