Image Cropping

Image cropping, the task of selecting the most visually appealing or informative portion of an image, is undergoing rapid advancement driven by the integration of large vision-language models (VLMs) and novel training strategies. Current research focuses on improving cropping accuracy and efficiency through techniques like prompt retrieval, iterative refinement, and pseudo-labeling to address data scarcity, while also incorporating user intent and design constraints. These improvements have significant implications for various applications, including content creation, document understanding, and image compression, by enabling more automated and aesthetically pleasing image manipulation.