The inconvenient truth of ground truth errors in automotive datasets and DNN-based detection

Assisted and automated driving functions will rely on machine learning algorithms, given their ability to cope with real-world variations, e.g. vehicles of different shapes, positions, colors, and so forth. Supervised learning needs annotated datasets, and several automotive datasets are available....

Full description

Saved in:
Bibliographic Details
Main Authors: Pak Hung Chan, Boda Li, Gabriele Baris, Qasim Sadiq, Valentina Donzella
Format: Article
Language:English
Published: Cambridge University Press 2024-01-01
Series:Data-Centric Engineering
Subjects:
Online Access:https://www.cambridge.org/core/product/identifier/S263267362400039X/type/journal_article
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846157687642914816
author Pak Hung Chan
Boda Li
Gabriele Baris
Qasim Sadiq
Valentina Donzella
author_facet Pak Hung Chan
Boda Li
Gabriele Baris
Qasim Sadiq
Valentina Donzella
author_sort Pak Hung Chan
collection DOAJ
description Assisted and automated driving functions will rely on machine learning algorithms, given their ability to cope with real-world variations, e.g. vehicles of different shapes, positions, colors, and so forth. Supervised learning needs annotated datasets, and several automotive datasets are available. However, these datasets are tremendous in volume, and labeling accuracy and quality can vary across different datasets and within dataset frames. Accurate and appropriate ground truth is especially important for automotive, as “incomplete” or “incorrect” learning can negatively impact vehicle safety when these neural networks are deployed. This work investigates the ground truth quality of widely adopted automotive datasets, including a detailed analysis of KITTI MoSeg. According to the identified and classified errors in the annotations of different automotive datasets, this article provides three different criteria collections for producing improved annotations. These criteria are enforceable and applicable to a wide variety of datasets. The three annotations sets are created to (i) remove dubious cases; (ii) annotate to the best of human visual system; and (iii) remove clear erroneous BBs. KITTI MoSeg has been reannotated three times according to the specified criteria, and three state-of-the-art deep neural network object detectors are used to evaluate them. The results clearly show that network performance is affected by ground truth variations, and removing clear errors is beneficial for predicting real-world objects only for some networks. The relabeled datasets still present some cases with “arbitrary”/“controversial” annotations, and therefore, this work concludes with some guidelines related to dataset annotation, metadata/sublabels, and specific automotive use cases.
format Article
id doaj-art-bbd861b1714c45b5977d7db5a19f931e
institution Kabale University
issn 2632-6736
language English
publishDate 2024-01-01
publisher Cambridge University Press
record_format Article
series Data-Centric Engineering
spelling doaj-art-bbd861b1714c45b5977d7db5a19f931e2024-11-25T06:27:06ZengCambridge University PressData-Centric Engineering2632-67362024-01-01510.1017/dce.2024.39The inconvenient truth of ground truth errors in automotive datasets and DNN-based detectionPak Hung Chan0https://orcid.org/0000-0003-1705-5430Boda Li1Gabriele Baris2Qasim Sadiq3Valentina Donzella4WMG, University of Warwick, Coventry, UKWMG, University of Warwick, Coventry, UKWMG, University of Warwick, Coventry, UKWMG, University of Warwick, Coventry, UKWMG, University of Warwick, Coventry, UKAssisted and automated driving functions will rely on machine learning algorithms, given their ability to cope with real-world variations, e.g. vehicles of different shapes, positions, colors, and so forth. Supervised learning needs annotated datasets, and several automotive datasets are available. However, these datasets are tremendous in volume, and labeling accuracy and quality can vary across different datasets and within dataset frames. Accurate and appropriate ground truth is especially important for automotive, as “incomplete” or “incorrect” learning can negatively impact vehicle safety when these neural networks are deployed. This work investigates the ground truth quality of widely adopted automotive datasets, including a detailed analysis of KITTI MoSeg. According to the identified and classified errors in the annotations of different automotive datasets, this article provides three different criteria collections for producing improved annotations. These criteria are enforceable and applicable to a wide variety of datasets. The three annotations sets are created to (i) remove dubious cases; (ii) annotate to the best of human visual system; and (iii) remove clear erroneous BBs. KITTI MoSeg has been reannotated three times according to the specified criteria, and three state-of-the-art deep neural network object detectors are used to evaluate them. The results clearly show that network performance is affected by ground truth variations, and removing clear errors is beneficial for predicting real-world objects only for some networks. The relabeled datasets still present some cases with “arbitrary”/“controversial” annotations, and therefore, this work concludes with some guidelines related to dataset annotation, metadata/sublabels, and specific automotive use cases.https://www.cambridge.org/core/product/identifier/S263267362400039X/type/journal_articlemachine learningautomated vehiclesautomotive datasetlabeling
spellingShingle Pak Hung Chan
Boda Li
Gabriele Baris
Qasim Sadiq
Valentina Donzella
The inconvenient truth of ground truth errors in automotive datasets and DNN-based detection
Data-Centric Engineering
machine learning
automated vehicles
automotive dataset
labeling
title The inconvenient truth of ground truth errors in automotive datasets and DNN-based detection
title_full The inconvenient truth of ground truth errors in automotive datasets and DNN-based detection
title_fullStr The inconvenient truth of ground truth errors in automotive datasets and DNN-based detection
title_full_unstemmed The inconvenient truth of ground truth errors in automotive datasets and DNN-based detection
title_short The inconvenient truth of ground truth errors in automotive datasets and DNN-based detection
title_sort inconvenient truth of ground truth errors in automotive datasets and dnn based detection
topic machine learning
automated vehicles
automotive dataset
labeling
url https://www.cambridge.org/core/product/identifier/S263267362400039X/type/journal_article
work_keys_str_mv AT pakhungchan theinconvenienttruthofgroundtrutherrorsinautomotivedatasetsanddnnbaseddetection
AT bodali theinconvenienttruthofgroundtrutherrorsinautomotivedatasetsanddnnbaseddetection
AT gabrielebaris theinconvenienttruthofgroundtrutherrorsinautomotivedatasetsanddnnbaseddetection
AT qasimsadiq theinconvenienttruthofgroundtrutherrorsinautomotivedatasetsanddnnbaseddetection
AT valentinadonzella theinconvenienttruthofgroundtrutherrorsinautomotivedatasetsanddnnbaseddetection
AT pakhungchan inconvenienttruthofgroundtrutherrorsinautomotivedatasetsanddnnbaseddetection
AT bodali inconvenienttruthofgroundtrutherrorsinautomotivedatasetsanddnnbaseddetection
AT gabrielebaris inconvenienttruthofgroundtrutherrorsinautomotivedatasetsanddnnbaseddetection
AT qasimsadiq inconvenienttruthofgroundtrutherrorsinautomotivedatasetsanddnnbaseddetection
AT valentinadonzella inconvenienttruthofgroundtrutherrorsinautomotivedatasetsanddnnbaseddetection