MNIST, NYU-Depth-V2

本研究涉及两个主要数据集：MNIST和NYU-Depth-V2。MNIST是一个包含60,000个手写数字图像的数据集，常用于图像分类问题。NYU-Depth-V2则包含RGB和深度图像，用于室内场景分析。研究中，MNIST数据集通过添加噪声和调整大小进行了定制化处理，而NYU-Depth-V2则被手动标记用于场景分类。这两个数据集的应用旨在通过深度估计辅助任务提升图像分类的准确性和鲁棒性，特别是在数据量有限和存在噪声的情况下。

This study involves two primary datasets: MNIST and NYU-Depth-V2. MNIST is a dataset containing 60,000 handwritten digit images, which is widely used for image classification tasks. NYU-Depth-V2, on the other hand, includes RGB and depth images for indoor scene analysis. In this research, the MNIST dataset was customized via noise addition and resizing operations, while NYU-Depth-V2 was manually annotated for scene classification. The application of these two datasets aims to enhance the accuracy and robustness of image classification through the auxiliary depth estimation task, especially in scenarios with limited data volume and existing noise.