Image-Bot: Everyday/Industrial Objects

Pictures of everyday and industrial objects in front of a greenscreen as well as masked and augmented images thereof. This dataset can be used to generate a larger synthetic training dataset for object detection and image segmentation by blending the masked images with different background images. The masked and augmented images consist of a pair of foreground image and mask, whereas the mask can be used to generate the labeling for the image segmentation dataset. This dataset was created as a prove of concept for and by the Image-Bot, which is described under https://github.com/FraunhoferIAO/Image-Bot as well as in the associated paper. The Image-Bot also offers the algorithms to blend the images of this dataset, with different background images. Just start with step 3 of the Image-Bot and name your project and object according to the objects in this dataset. Furthermore, this dataset contains evaluation pictures for each object, to test algorithms trained with the synthetic dataset from the masked objects. If you use this dataset, please cite it via its DOI and via the original paper about the Image-Bot.

搜集汇总

背景与挑战

背景概述

该数据集名为'Image-Bot: Everyday/Industrial Objects'，发布于2022年，包含日常和工业对象在绿幕前的图片及其掩码和增强图像，用于生成合成训练数据以支持对象检测和图像分割任务。数据集由Image-Bot工具创建，作为概念验证，提供了多种对象类别（如香蕉、瓶子、盒子等），并遵循CC BY 4.0许可证，适用于制造和日常活动相关的研究。