Multimodal-WCE-1 (MM-WCE-1) and Multimodal-WCE-2 (MM-WCE-2)

该数据集在先进的渲染技术支持下，于虚拟胶囊内窥镜环境中收集而成。它包含了来自单目无线胶囊内窥镜的振动信号和视觉数据，并附有标注的振动强度、真实深度图以及自运动数据。该数据集旨在通过多模态学习减轻振动噪声，并在真实场景中验证了其卓越的泛化能力。数据集规模涵盖多个视频序列，分别用于训练、验证和测试。其任务旨在进行深度估计和自运动估计。

This dataset was collected in a virtual capsule endoscopy environment powered by advanced rendering techniques. It contains vibration signals and visual data from monocular wireless capsule endoscopes, along with annotated vibration intensity, ground-truth depth maps and egomotion data. This dataset aims to mitigate vibration noise through multimodal learning, and its excellent generalization ability has been validated in real-world scenarios. The dataset comprises multiple video sequences, which are separately used for training, validation and testing. Its targeted tasks are depth estimation and egomotion estimation.