For some academic work, I am training an AI object detection algorithm (TensorFlow models) to look for specific objects (plants, in my case). I am taking photos with a hand-held camera, and am having to tile the images into smaller images (1024 x 1024 pixels) due to computer memory limitations. This will make the objects appear "bigger" in these tiled images.
As can be seen, the test images will be smaller than the trained images, so would this cause any issues in terms of accuracy when evaluating the model? How could I train my model(s) to detect these smaller images? (They are not GIS-tagged). I could always crop the test images to the same dimension, but would like to avoid doing this if at all possible. Thinking about using Scaled YOLOv4, would this work?