WaspMOT
收藏资源简介:
WaspMOT是由法国国家信息与自动化研究所、法国国家农业食品与环境研究院等机构联合创建的长时多目标跟踪基准数据集,专注于在实验室控制环境下对赤眼蜂进行长期身份一致性追踪。该数据集包含10个视频序列,总计约12万帧图像和256.9万标注实例,每个序列持续约8分钟(12,000帧),以25 FPS录制,分辨率达3840×2160像素,数据来源于固定俯视摄像机记录的实验竞技场。数据集创建过程涉及在气候控制实验室中录制赤眼蜂行为,并采用MOTChallenge标准格式进行密集边界框标注,同时提供完美检测以隔离关联性能评估。该数据集主要应用于计算机视觉领域的长时多目标跟踪算法评测,旨在解决现有基准在长期身份保持能力评估上的不足,特别针对生态监测中个体行为分析所需的身份一致性挑战。
WaspMOT is a long-term multi-object tracking benchmark dataset jointly created by institutions including the French National Institute for Computer Science and Applied Mathematics (INRIA) and the French National Institute for Agriculture, Food, and Environment (INRAE), focusing on long-term identity-consistent tracking of Trichogramma wasps in a laboratory-controlled environment. This dataset includes 10 video sequences, totaling approximately 120,000 frames and 2.569 million annotated instances. Each sequence lasts around 8 minutes (12,000 frames), recorded at 25 FPS with a resolution of 3840×2160 pixels, and the data is sourced from an experimental arena captured by a fixed overhead camera. The dataset creation process involves recording the behavior of Trichogramma wasps in a climate-controlled laboratory, performing dense bounding box annotations in the MOTChallenge standard format, and providing perfect detections to isolate the evaluation of tracking association performance. This dataset is primarily applied to the evaluation of long-term multi-object tracking algorithms in the computer vision field, aiming to address the shortcomings of existing benchmarks in evaluating long-term identity retention capabilities, particularly targeting the identity consistency challenge required for individual behavior analysis in ecological monitoring.




