Change Detection Meets Visual Question Answering

The Earth's surface is continually changing, and identifying changes plays an important role in urban planning sustainability. Although change detection techniques have been successfully developed for many years, these are still limited to experts facilitators related fields. In order provide every user with flexible access information help them better understand land-cover changes, we introduce a novel task: detection-based visual question answering (CDVQA) on multi-temporal aerial images. particular, images can be queried obtain high level change-based according content between two input We first build CDVQA dataset including image-question-answer triplets using automatic question-answer generation method. Then, baseline framework devised this work, it contains four parts: feature encoding, fusion, multi-modal answer prediction. addition, also enhancing module aiming at incorporating more change-related information. Finally, effects of different backbones fusion strategies studied the performance task. experimental results useful insights developing models, which future research