FSS-1000 小样本图片分割数据集

FSS-1000 是一个小样本分割的数据集，包含 1000 个类。该数据集探究仅通过 5 张人工标注的图像，训练模型完成图像识别任务。数据集包含大量在以前的数据集中，从未出现过或标注过的物体，如日常小物件、商品、卡通人物、 logo 等。

FSS-1000 is a few-shot segmentation dataset encompassing 1000 classes. This dataset is designed to explore training models to complete image recognition tasks using only 5 manually annotated images. It contains a large number of objects that have never appeared or been annotated in prior datasets, such as daily small items, commercial goods, cartoon characters, logos, and so on.

搜集汇总

背景与挑战

背景概述

FSS-1000是一个小样本图像分割数据集，涵盖1000个类别，旨在利用仅5张人工标注的图像来训练模型进行图像识别。该数据集包含许多以往数据集中未出现或标注的物体，例如日常小物件、商品、卡通人物和logo。