2026-10-13T17:59:58Z https://www.scidb.cn/oai

10.57760/sciencedb.j00240.00113 2026-03-26T14:22:07Z

2026-03-26 AI-generated plane image doi:10.57760/sciencedb.j00240.00113 en This dataset is constructed through an automated pipeline based on large language models consisting of three stages. First, Kimi 2.5 is employed to simulate the reasoning process of an aviation photography expert, transforming six dimensions of structured parameters—such as aircraft type, viewpoint, and environment—into diverse English prompts. Second, these prompts are fed into two advanced open-source image generation models, Qwen-Image-2512 and Z-Image-Turbo, for image synthesis. Finally, an integrated pipeline with a multi-threaded asynchronous producer–consumer mechanism is used to complete image collection and automatic retry.The dataset contains a total of 3,200 generated aircraft images, with each model (Qwen-Image and Z-Image) producing 1,600 images. The data are generated based on combinational parameter spaces and include the following attributes:Viewpoint: front view, side view, top view, bottom view Distance: near, far Weather: sunny, rainy, cloudy, overcast, foggy Background: sky, mountains, forest, ocean, city Quantity: 1 aircraft, 3 aircraft Aircraft Type: U.S. F-35 fighter jet, Russian Su-57 fighter jet, U.S. RQ-180 unmanned reconnaissance aircraft, Russian S350M unmanned reconnaissance aircraft All generated images are uniformly set to a resolution of 1664 × 928. AI-generated; plane images; diverse scenarios Peisong He Shuai Tang RESTRICTED dataset Science Data Bank