ETHZ Shape Classes

ETHZ Shape 数据集包含五种基于形状的不同类别的图像，这些图像是从 Flickr 和 Google Images 收集的。它提供的主要挑战是混乱、类内形状可变性和尺度变化。作者特意选择了几张图像，其中对象仅占图像的一小部分，并努力包含出现在广泛范围内的对象。这些对象大多是未被遮挡的，并且都是从大致相同的视点（侧面）拍摄的。

The ETHZ Shape Dataset contains images belonging to five distinct shape-based categories, which were collected from Flickr and Google Images. The primary challenges it presents include clutter, intra-class shape variability, and scale variation. The authors intentionally selected several images where the objects only occupy a small portion of the image frame, and strived to include objects spanning a wide range of scales. Most of these objects are unoccluded, and all were captured from roughly the same lateral viewpoint.

搜集汇总

背景与挑战

背景概述

ETHZ Shape Classes数据集包含五种基于形状的不同类别图像，主要挑战是混乱、类内形状可变性和尺度变化。该数据集由爱丁堡大学和卡昂诺曼底大学于2006年发布，适用于图像分类任务。