Understanding Using TensorFlow to Determine Objects

You mentioned object detection in the comment section, but
"This part in the image is not ""a dog"" but
" "If ""this image is a dog image"" is all you need, you don't need an object detection."

In this case, you will be using a convolutional neural network. Tutorials are available on TensorFlow.
https://www.tensorflow.org/tutorials/
Tutorials are a basic image classification problem called MNIST, so
For complex images like dogs and cats, I don't think the accuracy will improve.
We also have ready-made models, so if you understand them in the tutorial, please refer to the following.
https://www.tensorflow.org/tutorials/images/image_recognition

Problem: If you let them determine which image is not a cat or a dog, the only way to do this is to get one of the closest ones as a solution (I don't know how to get the rest)

"Regarding , I think it would be better to let them learn images that are not cats or dogs at the time of learning, and label them as ""others."""