ILSVRC-2012
收藏资源简介:
本研究中使用的数据集为ILSVRC-2012,由哈尔滨工业大学等机构创建,旨在通过自动数据增强技术提升图像识别任务的性能。该数据集通过网络获取了额外的1250万张图像,这些图像带有丰富的上下文信息,通过深度卷积神经网络(DCNN)进行自动标注。数据集的创建过程结合了网络的上下文信息和DCNN的视觉信息,以提高标注的准确性和数据集的丰富性。该数据集主要应用于图像识别领域,特别是通过自动增强现有数据集来接近预期误差,解决传统数据标注成本高和比较不公平的问题。
The dataset used in this study is ILSVRC-2012, developed by Harbin Institute of Technology and other institutions. It was constructed with the goal of enhancing the performance of image recognition tasks through automatic data augmentation techniques. The dataset collected an additional 12.5 million images from the web; these images carry rich contextual information and were automatically annotated using Deep Convolutional Neural Networks (DCNNs). The dataset's construction process combines the contextual information of these web-sourced images and the visual information processed by DCNNs, to improve annotation accuracy and enrich the dataset's diversity and comprehensiveness. This dataset is primarily applied in the field of image recognition, specifically to narrow the gap to the expected optimal error by automatically augmenting existing datasets, and to address the issues of high data annotation costs and unfair comparative evaluations in traditional data workflows.




