MarSyn

海事合成 (MarSyn) 数据集由使用Blender获得的25个不同的真实感视频序列组成，每个包含1000帧。它的25,000图像及其相应的34,000注释旨在模拟各种海洋场景和条件，例如晴天，多云和日落环境，近岸图像和水面反射。在模拟中部署了不同类型的船舶，这增加了数据集的多样性。这些船只的类型 (货物，军舰，捕鱼，快艇，救援筏等)，长度 (从3米一直到125米)，形状和颜色。模拟摄像机被放置并相应地移动到无人机上，无人机可以在150到1000米的高度捕获图像。

The Maritime Synthetic (MarSyn) dataset comprises 25 distinct photorealistic video sequences generated using Blender, with each sequence containing 1000 frames. A total of 25,000 images and their corresponding 34,000 annotations are designed to simulate diverse maritime scenarios and conditions, including clear-sky, cloudy, and sunset environments, near-shore imagery, and water surface reflections. Various types of vessels are deployed during the simulation to enhance the dataset's diversity. These vessels feature different categories (cargo ships, warships, fishing boats, speedboats, rescue rafts, etc.), with lengths ranging from 3 meters to 125 meters, as well as varied shapes and colors. The simulated cameras are mounted and moved correspondingly on drones, which can capture images at altitudes between 150 and 1000 meters.

搜集汇总

背景与挑战

背景概述

MarSyn是一个海事合成数据集，包含25,000张图像和34,000个注释，模拟了多种海洋场景和不同类型的船舶。数据集由系统和机器人研究所·海军研究中心于2022年发布，旨在提供多样化的海洋场景和船舶类型数据。