MammAlps
收藏资源简介:
MammAlps是由洛桑联邦理工学院创建的一个多模态、多视角野生动物行为监测数据集,包含在瑞士国家公园从9个摄像头陷阱收集的超过14小时的视频和音频,以及8.5小时的个体轨迹密集标注信息。该数据集基于6135个单一动物视频片段,提出了第一个使用音频、视频和场景分割图作为输入的分层、多模态动物行为识别基准,同时还提出了一个面向生态学的多视角、长期事件理解基准,旨在将长期生态事件总结为有意义的生态行为信息。
MammAlps is a multimodal, multi-view wildlife behavior monitoring dataset created by École Polytechnique Fédérale de Lausanne. It contains over 14 hours of video and audio data collected from 9 camera traps in Swiss National Parks, along with 8.5 hours of densely annotated individual trajectory information. Built on 6,135 single-animal video clips, this dataset introduces the first hierarchical, multimodal animal behavior recognition benchmark that uses audio, video and scene segmentation maps as inputs. Additionally, it proposes an ecology-oriented multi-view long-term event understanding benchmark, which aims to summarize long-term ecological events into meaningful ecological behavioral information.

- 1MammAlps: A multi-view video behavior monitoring dataset of wild mammals in the Swiss Alps洛桑联邦理工学院(Ecole Polytechnique Fédérale de Lausanne, EPFL) · 2025年



