Cut-and-Paste Dataset

Cut-and-Paste Dataset是由电子与电信研究所创建的一个用于对象实例检测的数据集。该数据集包含约6000张图像，通过从训练图像中剪切对象区域并将其粘贴到其他背景图像上生成。数据集的创建旨在解决在只有少量训练对象图像可用时训练对象实例检测器的挑战。通过这种方法，数据集能够模拟不同环境下的对象检测，从而提高检测器的泛化能力。该数据集特别适用于解决室内复杂环境中的对象检测问题，通过平衡前景和背景的域间隙，提高了检测精度。

The Cut-and-Paste Dataset is a dataset for object instance detection created by the Institute of Electronics and Telecommunications. It contains approximately 6,000 images generated by cropping object regions from training images and pasting them onto other background images. The dataset was developed to address the challenge of training object instance detectors when only a limited number of training object images are available. Through this approach, the dataset can simulate object detection scenarios in diverse environments, thereby enhancing the generalization performance of detectors. This dataset is particularly suitable for solving object detection problems in complex indoor environments, as it balances the domain gap between foreground and background, thus improving detection accuracy.