Image-Bot: Everyday/Industrial Objects

Published: 23 December 2021| Version 1 | DOI: 10.17632/4nn2w8rvx3.1
Contributors:
,
,

Description

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/MobilityInnovation/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 or via the original paper about the Image-Bot.

Files

Steps to reproduce

This dataset was created as a prove of concept for and by the Image-Bot, which is described under https://github.com/MobilityInnovation/Image-Bot as well as in the associated paper. Follow the procedure described therein to reproduce the dataset or create a similar one.

Institutions

Universitat Stuttgart Institut fur Arbeitswissenschaft und Technologiemanagement, Fraunhofer-Institut fur Arbeitswirtschaft und Organisation, Eberhard Karls Universitat Tubingen

Categories

Activities of Daily Living, Image Segmentation, Object Detection, Manufacturing, Synthetic Image

License