Venus: Benchmarking and Empowering Multimodal Large Language Models for Aesthetic Guidance and Cropping

The cropped image focuses on the architectural details of the building, emphasizing grandeur and symmetry. By removing the sky and lower portion, the composition becomes more balanced and centered, drawing attention to the columns and dome for a stronger visual impact. Venus (ours)：The composition appears somewhat lacking in a clear focal point and a sense of emotional depth, as the linear perspective of the benches holds potential but is not fully utilized. To enhance the overall visual impact, it is recommended to adjust the shooting angle at night, utilizing the circular ring as a framing element to guide the viewer's gaze toward the distant skyscrapers, thereby strengthening the visual focus and enhancing the sense of depth.

AesExpert：The clarity of this image is high, and the details are depicted vividly and accurately, with sharp edges. The color variety is abundant, and the combination is natural and harmonious, resulting in a pleasing visual impression. The composition adopts a central composition, highlighting the subject. Therefore, this image looks quite beautiful.

Re-captured Photo Reshoot Following Aesthetic Guidance Current MLLMs Task 2: Aesthetic Cropping (after capture) Re-captured Photo GPT-4o AesExpert Venus (ours) Fail to Identify Issues

GPT-4o: This photo effectively uses perspective and composition to create depth and interest. The leading lines of the wooden planks draw the viewer's eye toward the background, where tall buildings and trees form a dynamic contrast between urban and natural elements… The composition harmonizes these details into a balanced and engaging image.