Multi-Input data of different sizes

I am working on a multi (or mixed) input problem, whereby each row in my tabular dataset has two accompanying images, hence if I have a tabular data with 300 rows, I will have 600 images.

I created a custom dataset class for this problem, however because I have 600 images and 300 rows, my len method will return only 300 which means I’m basically discarding half my images. I’m wondering if I have the right idea or if someone can point me in the right direction. I was initially thinking of adding an extra argument to the dataset’s constructor (i.e image1 and image2), but I think this means I will have to create two CNN’s (one for each image) and I think this might be overkill for my problem.