How to create Tensorflow Dataset with input images and multiple target features in .csv

Solutions in the Tensorflow Tutorials are not very helpful for me, because they are focused on the one-label classification task. And I am learning how to make regression model with multiple target features for each image.

I have inputs as .png files on my disk and targets as a single .csv file. Each image have 3 corresponding integer target features.

/Project
   /Images
       0.png
       1.png
       2.png
       3.png
         ...

   /Targets
       targets.csv

I can load all images to the Dataset using 'image_dataset_from_directory':

data_dir = "/Project/Images"

inputs_ds = tensorflow.keras.utils.image_dataset_from_directory(
  data_dir,
  labels=None,
  color_mode="rgb",
  seed=1,
  image_size=(256, 256),
  batch_size=None)

Now I'm stuck on adding target features from .csv to the Dataset. How can I do it?