Letterboxing is a common image preprocessing method for size adaptation. It scales the image to the target size while preserving the original aspect ratio, then fills the image edges with fixed pixel values (usually black) to avoid stretching distortion and meet the fixed input size requirements of models, a classic preprocessing operation in object detection tasks.
