Automatic Bounding Box Annotation with Small Training Datasets for Industrial Manufacturing
In the past few years, object detection has attracted a lot of attention in the context of human–robot collaboration and Industry 5.0 due to enormous quality improvements in deep learning technologies. In many applications, object detection models have to be able to quickly adapt to a changing envir...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-02-01
|
| Series: | Micromachines |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-666X/14/2/442 |
| _version_ | 1797619222997630976 |
|---|---|
| author | Manuela Geiß Raphael Wagner Martin Baresch Josef Steiner Michael Zwick |
| author_facet | Manuela Geiß Raphael Wagner Martin Baresch Josef Steiner Michael Zwick |
| author_sort | Manuela Geiß |
| collection | DOAJ |
| description | In the past few years, object detection has attracted a lot of attention in the context of human–robot collaboration and Industry 5.0 due to enormous quality improvements in deep learning technologies. In many applications, object detection models have to be able to quickly adapt to a changing environment, i.e., to learn new objects. A crucial but challenging prerequisite for this is the automatic generation of new training data which currently still limits the broad application of object detection methods in industrial manufacturing. In this work, we discuss how to adapt state-of-the-art object detection methods for the task of automatic bounding box annotation in a use case where the background is homogeneous and the object’s label is provided by a human. We compare an adapted version of Faster R-CNN and the Scaled-YOLOv4-p5 architecture and show that both can be trained to distinguish unknown objects from a complex but homogeneous background using only a small amount of training data. In contrast to most other state-of-the-art methods for bounding box labeling, our proposed method neither requires human verification, a predefined set of classes, nor a very large manually annotated dataset. Our method outperforms the state-of-the-art, transformer-based object discovery method <i>LOST</i> on our simple fruits dataset by large margins. |
| first_indexed | 2024-03-11T08:23:58Z |
| format | Article |
| id | doaj.art-3bdaacb894654f15b73652f092a7e343 |
| institution | Directory Open Access Journal |
| issn | 2072-666X |
| language | English |
| last_indexed | 2024-03-11T08:23:58Z |
| publishDate | 2023-02-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Micromachines |
| spelling | doaj.art-3bdaacb894654f15b73652f092a7e3432023-11-16T22:12:18ZengMDPI AGMicromachines2072-666X2023-02-0114244210.3390/mi14020442Automatic Bounding Box Annotation with Small Training Datasets for Industrial ManufacturingManuela Geiß0Raphael Wagner1Martin Baresch2Josef Steiner3Michael Zwick4Software Competence Center Hagenberg GmbH, Softwarepark 32a, 4232 Hagenberg, AustriaSoftware Competence Center Hagenberg GmbH, Softwarepark 32a, 4232 Hagenberg, AustriaKEBA Group AG, Reindlstraße 51, 4040 Linz, AustriaKEBA Group AG, Reindlstraße 51, 4040 Linz, AustriaSoftware Competence Center Hagenberg GmbH, Softwarepark 32a, 4232 Hagenberg, AustriaIn the past few years, object detection has attracted a lot of attention in the context of human–robot collaboration and Industry 5.0 due to enormous quality improvements in deep learning technologies. In many applications, object detection models have to be able to quickly adapt to a changing environment, i.e., to learn new objects. A crucial but challenging prerequisite for this is the automatic generation of new training data which currently still limits the broad application of object detection methods in industrial manufacturing. In this work, we discuss how to adapt state-of-the-art object detection methods for the task of automatic bounding box annotation in a use case where the background is homogeneous and the object’s label is provided by a human. We compare an adapted version of Faster R-CNN and the Scaled-YOLOv4-p5 architecture and show that both can be trained to distinguish unknown objects from a complex but homogeneous background using only a small amount of training data. In contrast to most other state-of-the-art methods for bounding box labeling, our proposed method neither requires human verification, a predefined set of classes, nor a very large manually annotated dataset. Our method outperforms the state-of-the-art, transformer-based object discovery method <i>LOST</i> on our simple fruits dataset by large margins.https://www.mdpi.com/2072-666X/14/2/442automatic object annotationimage annotationobject detectionAutoMLdeep learningIndustry 5.0 |
| spellingShingle | Manuela Geiß Raphael Wagner Martin Baresch Josef Steiner Michael Zwick Automatic Bounding Box Annotation with Small Training Datasets for Industrial Manufacturing Micromachines automatic object annotation image annotation object detection AutoML deep learning Industry 5.0 |
| title | Automatic Bounding Box Annotation with Small Training Datasets for Industrial Manufacturing |
| title_full | Automatic Bounding Box Annotation with Small Training Datasets for Industrial Manufacturing |
| title_fullStr | Automatic Bounding Box Annotation with Small Training Datasets for Industrial Manufacturing |
| title_full_unstemmed | Automatic Bounding Box Annotation with Small Training Datasets for Industrial Manufacturing |
| title_short | Automatic Bounding Box Annotation with Small Training Datasets for Industrial Manufacturing |
| title_sort | automatic bounding box annotation with small training datasets for industrial manufacturing |
| topic | automatic object annotation image annotation object detection AutoML deep learning Industry 5.0 |
| url | https://www.mdpi.com/2072-666X/14/2/442 |
| work_keys_str_mv | AT manuelageiß automaticboundingboxannotationwithsmalltrainingdatasetsforindustrialmanufacturing AT raphaelwagner automaticboundingboxannotationwithsmalltrainingdatasetsforindustrialmanufacturing AT martinbaresch automaticboundingboxannotationwithsmalltrainingdatasetsforindustrialmanufacturing AT josefsteiner automaticboundingboxannotationwithsmalltrainingdatasetsforindustrialmanufacturing AT michaelzwick automaticboundingboxannotationwithsmalltrainingdatasetsforindustrialmanufacturing |