Train test image quality difference


I am currently training my age-gender estimation model on images from various datasets ( with different image quality if it makes sense) and will be testing it on images obtained from either a webcam or CCTV.

I plan to add image quality enhancements like increasing sharpness and contrast for the test set. I was wondering if there are any similar experiments performed and how the results were.

Intuitively, I understand that the model should have no problem predicting in better quality images and would like to check more sources.

Thank You