Hey so it's been a while since I've worked with DL4J so I would advice to check out the documentation here:

https://deeplearning4j.konduit.ai/v/en-1.0.0-beta7/android/image-classification

And also it's better to have images of same hight and width as answered here: https://www.quora.com/To-what-resolution-should-I-resize-my-images-to-use-as-training-dataset-for-deep-learning