MultiFloodSynth
收藏资源简介:
MultiFloodSynth是由韩国中央大学创建的一个合成数据集,旨在为洪水灾害检测提供高质量的训练数据。该数据集通过控制多个参数来模拟不同级别的洪水情况,并生成相应的虚拟场景。数据集包含70,117张图像,其中有14,593张洪水图像和55,524张非洪水图像,提供了9种类型的注释,包括语义/实例/细粒度分割、2D/3D边界框等。该数据集可应用于各种计算机视觉任务,特别是在对象定位的洪水级别识别方面表现出色。
MultiFloodSynth is a synthetic dataset developed by Chung-Ang University, Republic of Korea, aimed at providing high-quality training data for flood disaster detection. This dataset simulates flood scenarios with varying severity levels by controlling multiple parameters, and generates corresponding virtual scenes. The dataset contains 70,117 images in total, including 14,593 flood images and 55,524 non-flood images. It provides nine types of annotations, such as semantic, instance and fine-grained segmentation, as well as 2D and 3D bounding boxes, among others. This dataset can be applied to various computer vision tasks, and performs particularly well in flood level recognition for object localization.

- 1MultiFloodSynth: Multi-Annotated Flood Synthetic Dataset Generation韩国中央大学 · 2025年



