CDLT
收藏资源简介:
CDLT数据集由华中科技大学电子信息与通信学院创建,专注于细粒度视觉分类中的概念漂移和长尾分布问题。该数据集包含11,195张图像,涵盖250个实例,跨越47个月,反映了自然环境中的真实变化。数据收集过程涉及众包工人和领域专家,确保了数据的质量和多样性。CDLT数据集旨在解决现有数据集在实际应用中可能遇到的性能问题,特别是在处理随时间变化的实例特征和类别不平衡时。该数据集的应用领域包括但不限于生物识别、环境监测和智能交互系统。
The CDLT dataset was developed by the School of Electronic Information and Communications, Huazhong University of Science and Technology, focusing on the issues of concept drift and long-tailed distribution in fine-grained visual classification. This dataset consists of 11,195 images covering 250 instances, spanning 47 months, which reflects real-world changes in natural environments. The data collection process involved crowdsourced workers and domain experts to guarantee the quality and diversity of the dataset. The CDLT dataset aims to address the performance challenges that existing datasets may face in real-world applications, particularly when handling temporally varying instance features and class imbalance. Application scenarios of this dataset include but are not limited to biometrics, environmental monitoring, and intelligent interactive systems.




