Exploring the Impact of Synthetic Data Generation on Texture-based Image Classification Tasks

Simple item page
dc.contributor.authorYordanov, Borislaven_US
dc.contributor.authorHarvey, Carloen_US
dc.contributor.authorWilliams, Ianen_US
dc.contributor.authorAshley, Craigen_US
dc.contributor.authorFairbrass, Paulen_US
dc.contributor.editorPelechano, Nuriaen_US
dc.contributor.editorLiarokapis, Fotisen_US
dc.contributor.editorRohmer, Damienen_US
dc.contributor.editorAsadipour, Alien_US
dc.date.accessioned2023-10-02T08:17:29Z
dc.date.available2023-10-02T08:17:29Z
dc.date.issued2023
dc.description.abstractIn this study, we introduce a novel pipeline for synthetic data generation of textured surfaces, motivated by the limitations of conventional methods such as Generative Adversarial Networks (GANs) and Computer-Aided Design (CAD) models in our specific context. We also investigate the pipeline's role in an image classification task. The primary objective is to determine the impact of synthetic data generated by our pipeline on classification performance. Using EfficientNetV2-S as our image classifier and a dataset of three texture classes, we find that synthetic data can significantly enhance classification performance when the amount of real data is scarce, corroborating previous research. However, we also observe that the balance between synthetic and real data is crucial, as excessive synthetic data can negatively impact performance when sufficient real data is available. We theorize that this might stem from imperfections in the synthetic data generation process that distort fine details essential for accurate classification, and propose possible improvements to the synthetic data generation pipeline. Furthermore, we acknowledge the potential limitations of our study and provide several promising avenues for future research. This work illuminates the advantages and potential drawbacks of synthetic data in image classification tasks, emphasizing the importance of high-quality, realistic synthetic data that complements, rather than undermines, the use of real data.en_US
dc.description.sectionheadersApplications in Digital Storytelling and Experience
dc.description.seriesinformationInternational Conference on Interactive Media, Smart Systems and Emerging Technologies (IMET)
dc.identifier.doi10.2312/imet.20231264
dc.identifier.isbn978-3-03868-233-2
dc.identifier.pages101-109
dc.identifier.pages9 pages
dc.identifier.urihttps://doi.org/10.2312/imet.20231264
dc.identifier.urihttps://diglib.eg.org:443/handle/10.2312/imet20231264
dc.publisherThe Eurographics Associationen_US
dc.rightsAttribution 4.0 International License
dc.rights.urihttps://creativecommons.org/licenses/by/4.0/
dc.subjectKeywords: Synthetic Data, Image Classification, Textured Surface CCS Concepts: Computing methodologies -> Object recognition; Appearance and texture representations; Image representations; Supervised learning by classification; Computer graphics
dc.subjectSynthetic Data
dc.subjectImage Classification
dc.subjectTextured Surface CCS Concepts
dc.subjectComputing methodologies
dc.subjectObject recognition
dc.subjectAppearance and texture representations
dc.subjectImage representations
dc.subjectSupervised learning by classification
dc.subjectComputer graphics
dc.titleExploring the Impact of Synthetic Data Generation on Texture-based Image Classification Tasksen_US
Files
Original bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
101-109.pdf
Size:
18.04 MB
Format:
Adobe Portable Document Format
Download
Collections
IMET 2023: International Conference on Interactive Media, Smart Systems and Emerging Technologies