KinaviScene100

KinaviScene100 is a synchronized multimodal dataset designed for inland navigation research, capturing real-world ship operations in the Wuhu Yangtze River Bridge waterway, China. The dataset comprises 100 navigation scenarios across 7 typical encounter types (overtaking, head-on, turning, bridge passing, crossing, berthing, and stopping), with a total duration of 167.86 minutes. Each scenario includes synchronized 360° panoramic video from 6-7 cameras (650 files, 84.33 GB), bridge audio recordings of crew communications and VHF radio (100 files, 1.66 GB), surrounding vessel AIS data (309 unique ships), own-ship GNSS positioning (9,258 data points), and IMU motion data (7,693 data points). All data streams are time-aligned to UTC with sub-100ms precision, enabling cross-modal analysis of crew behavior, navigation decision-making, and traffic situation awareness. This dataset supports research in maritime human factors, speech recognition for navigation commands, ship detection and tracking, multimodal data fusion, and autonomous ship development.

KinaviScene100是一款专为内河航运研究开发的同步多模态数据集（multimodal dataset），采集自中国芜湖长江大桥水域的真实船舶作业场景。该数据集包含7类典型会遇场景下共计100个航行工况，总时长167.86分钟，7类典型会遇场景分别为追越、对遇、转向、过桥、交叉会遇、靠泊与停船。每个工况均包含由6-7台摄像机同步采集的360°全景视频（共计650个文件，总容量84.33GB）、桥载船员通信与甚高频（VHF, Very High Frequency）无线电音频录音（100个文件，总容量1.66GB）、周边船舶自动识别系统（AIS, Automatic Identification System）数据（涉及309艘独特船舶）、本船全球导航卫星系统（GNSS, Global Navigation Satellite System）定位数据（9258条记录）以及惯性测量单元（IMU, Inertial Measurement Unit）运动数据（7693条记录）。所有数据流均以低于100毫秒的精度对齐至协调世界时（UTC, Coordinated Universal Time），可支持针对船员行为、航行决策与交通态势感知的跨模态分析。本数据集可应用于海事人因工程、航行指令语音识别、船舶检测与跟踪、多模态数据融合以及自主船舶研发等相关研究方向。