Unlabeled wide-angle portraits correction dataset

按照现有方法，我们对广角数据集进行了广泛的实验，用5种不同的智能手机捕获。训练数据集包含测试数据集中的5，000多个图像和129。为数据集中的每个图像提供了多种标签，其中包含面罩，校正流程图和地标。此外，我们通过4种不同的智能手机 (包括三星Note 10，小米11，vivo X23和vivo iQOO) 收集了超过5,000张图像，作为未标记的集合。

Following existing experimental methodologies, we conducted extensive experiments on the wide-angle dataset, which was captured using five distinct smartphone models. The training dataset includes over 5,000 images and 129 from the test dataset. Multiple labels are provided for each image in the dataset, including masks, correction flowcharts, and landmarks. Additionally, we collected more than 5,000 images using four different smartphone models, namely Samsung Note 10, Xiaomi 11, vivo X23, and vivo iQOO, to serve as the unlabeled dataset.

搜集汇总

背景与挑战

背景概述

该数据集专注于广角肖像校正，包含训练集和测试集，其中训练集有超过5,000张图像，测试集有129张图像，每张图像均提供面罩、校正流程图和地标等标签。此外，还通过4种智能手机收集了超过5,000张未标记图像，由多所机构于2022年发布。