CAMO(Camouflaged Object)

伪装对象 (CAMO) 数据集，专为伪装对象分割任务而设计。我们关注两类，即自然伪装的物体和人工伪装的物体，它们通常分别对应于现实世界中的动物和人类。 伪装对象图像由 1250 张图像组成（训练集 1000 张图像，测试集 250 张图像）。从 MS-COCO 数据集中收集非伪装对象图像（训练集 1000 张图像，测试集 250 张图像）。 CAMO 具有客观性掩码真实性。The CAMO (Camouflaged Object) dataset is specifically designed for the task of camouflaged object segmentation. This dataset focuses on two categories: naturally camouflaged objects and artificially camouflaged objects, which typically correspond to animals and humans in real-world scenarios respectively. A total of 1250 camouflaged object images are included, with 1000 images allocated to the training set and 250 to the test set. Non-camouflaged object images are collected from the MS-COCO dataset, with 1000 images for the training set and 250 images for the test set. The CAMO dataset provides objective ground-truth masks.

搜集汇总

背景与挑战

背景概述

CAMO(Camouflaged Object)数据集是一个专为伪装对象分割任务设计的图像数据集，包含自然伪装和人工伪装两类对象，分别对应现实世界中的动物和人类。数据集由1250张伪装对象图像（训练集1000张，测试集250张）和从MS-COCO收集的同等数量的非伪装对象图像组成，具有客观性掩码真实性，适用于计算机视觉研究。该数据集于2019年由多个大学机构发布，采用CC BY-NC-SA 3.0许可证。