MOD20
收藏资源简介:
MOD20数据集是由Defence Science and Technology Group创建的,专注于户外多视角人体动作识别。该数据集包含2324个视频片段,总计503086帧,覆盖20种动态人体动作类别。数据来源于YouTube和自主无人机拍摄,视频分辨率为720×720,确保原始比例不变。创建过程涉及从多个视角收集视频,并进行精细的标注工作。MOD20数据集适用于动作识别、监控和态势感知等研究领域,旨在解决户外复杂环境中的人体动作识别问题。
The MOD20 dataset was developed by the Defence Science and Technology Group, focusing on multi-view human action recognition in outdoor environments. This dataset comprises 2,324 video clips, amounting to a total of 503,086 frames, and encompasses 20 categories of dynamic human actions. The data is sourced from YouTube and footage captured by autonomous drones, with a resolution of 720×720 to preserve the original aspect ratio. The dataset curation process involved collecting videos from multiple perspectives and conducting meticulous annotation work. The MOD20 dataset is applicable to research fields including action recognition, surveillance, and situation awareness, and aims to address the challenges of human action recognition in complex outdoor environments.

- 1A Multi-viewpoint Outdoor Dataset for Human Action RecognitionDefence Science and Technology Group · 2021年



