What is the effect of square input image scaling on the performance of deep learning models?

  1. From the conclusion, is it influential or not?:thinking:

  2. How much is the impact? What factors make an impression?:thinking:

  3. What is the mathematical theory derivation?:thinking:

Here is a similar topic, but it’s NOT a square image input.:expressionless: