BIPED (Barcelona Images for Perceptual Edge Detection)

详细信息 它包含 250 张户外图像，每张 1280 美元\乘以 720 美元像素。这些图像已由计算机视觉领域的专家仔细注释，因此没有考虑冗余。尽管如此，所有结果都经过了多次交叉检查，以纠正一个受试者可能出现的错误或错误的边缘。该数据集可作为评估边缘检测算法的基准公开获得。该数据集的生成是由于缺乏边缘检测数据集，实际上，只有一个公开的数据集可用于 2016 年发布的边缘检测任务（MDBD：Multicue Dataset for Boundary Detection——边缘检测的子集）。 BIPED 图像中边缘级别注释的细节级别可以通过 GT 来了解，参见上图。 BIPED 数据集有 250 张高清图像。这些图像已经被分割用于训练和测试。 200 个用于训练，50 个用于测试。版本 当前版本是第二个。Detailed Information This dataset contains 250 outdoor images, each with a resolution of 1280 × 720 pixels. All images have been carefully annotated by experts in the field of computer vision, with no redundant annotations included. Nevertheless, all annotations have undergone multiple cross-checks to correct potential errors or inaccurate edge labels from individual annotators. This dataset is publicly available as a benchmark for evaluating edge detection algorithms. This dataset was developed to address the lack of existing edge detection datasets. In fact, only one publicly available dataset existed for the edge detection task released in 2016: MDBD (Multicue Dataset for Boundary Detection, a subset for edge detection). The level of detail for edge-level annotations in BIPED images can be understood via the ground truth (GT); please refer to the figure above. The BIPED dataset consists of 250 high-definition images, which have been split into training and test sets: 200 images for training and 50 for testing. Version The current version is Version 2.

搜集汇总

背景与挑战

背景概述

BIPED是一个包含250张高清户外图像的边缘检测数据集，其中200张用于训练，50张用于测试。该数据集由专家标注，旨在为边缘检测算法提供评估基准，并解决该领域数据集缺乏的问题。