What would happen if the image size of training dataset is smaller than Object Detection Neural Network input image size?

Question

The scenario is i want to train a object detection model based on yolov5, the default input image size of yolov5 is 640×640, but my dataset have some images whose image size