CASIA-B
收藏资源简介:
Dataset B * (CASIA-B *) 是由Liang等人处理的CASIA-B的重新分割版本。CASIA-B * 的额外导入归因于CASIA-B用于生成轮廓数据的背景减法算法,该算法往往会产生很多噪声,并且已经过时了。当今世界应用。他们使用最新的预处理策略来重新分割原始视频,即深度行人轨迹和分割算法。结果,CASIA-B * 由裁剪的RGB图像,二进制轮廓和获得的边界框的高宽比组成。
Dataset B* (CASIA-B*) is a re-segmented version of the original CASIA-B dataset processed by Liang et al. The additional inclusion of CASIA-B* is attributed to the fact that the background subtraction algorithm used by the original CASIA-B to generate silhouette data tends to produce excessive noise and is outdated for modern real-world applications. The researchers adopted the latest preprocessing strategies, specifically deep pedestrian tracking and segmentation algorithms, to re-segment the original video data. Consequently, CASIA-B* consists of cropped RGB images, binary silhouettes, and the aspect ratios of the obtained bounding boxes.

- CASIA-B数据集首次发表,由中科院自动化研究所(CASIA)发布,旨在为步态识别研究提供标准化的数据资源。
- CASIA-B数据集首次应用于步态识别算法的研究与评估,成为该领域的重要基准数据集之一。
- CASIA-B数据集被广泛应用于国际步态识别竞赛中,进一步提升了其在学术界的影响力。
- 随着深度学习技术的发展,CASIA-B数据集开始被用于训练和验证基于深度学习的步态识别模型。
- CASIA-B数据集继续作为步态识别研究的重要资源,支持了多项前沿技术的开发与应用。
- 1CASIA Gait Database (GAITDB)Institute of Automation, Chinese Academy of Sciences · 2005年
- 2Gait Recognition Using a View Transformation Model in the Frequency DomainInstitute of Automation, Chinese Academy of Sciences · 2006年
- 3Gait Recognition by Motion Context from a Single ImageUniversity of Oulu · 2019年
- 4Gait Recognition Using Deep Learning: A SurveyUniversity of Surrey · 2020年
- 5Gait Recognition Using Deep Convolutional Neural NetworksUniversity of California, Irvine · 2018年



