A Study on Classifying Automotive Pedal Dashcam Images Using Transfer Learning 


Vol. 14,  No. 2, pp. 104-112, Feb.  2025
https://doi.org/10.3745/TKIPS.2025.14.2.104


PDF
  Abstract

This study comprehensively evaluates the performance of six major transfer learning models (VGG, ResNet, EfficientNet, Inception, DenseNet, and MobileNet) to classify the operation of the driver's brake and accelerator pedals using vehicle dashboard camera image data. The dataset consists of over 24,000 images collected in various driving environments and conditions, with each image clearly labeled to indicate the driver's pedal operation status. Through systematic experiments, the strengths and weaknesses of each model were analyzed, revealing that each transfer learning model exhibited different patterns based on their characteristics and maintained high performance even with complex image data. The results of this study contribute to a deeper understanding of the effectiveness of various transfer learning architectures and provide practical guidance for selecting the optimal model for driving behavior analysis. This research serves as a foundational resource for analyzing vehicle sudden acceleration incidents and is expected to contribute to the advancement of vehicle safety systems while enhancing image data processing in automotive environments.

  Statistics
  Cite this article

[IEEE Style]

J. Lim and H. Choi, "A Study on Classifying Automotive Pedal Dashcam Images Using Transfer Learning," The Transactions of the Korea Information Processing Society, vol. 14, no. 2, pp. 104-112, 2025. DOI: https://doi.org/10.3745/TKIPS.2025.14.2.104.

[ACM Style]

JongBeom Lim and HeeSeok Choi. 2025. A Study on Classifying Automotive Pedal Dashcam Images Using Transfer Learning. The Transactions of the Korea Information Processing Society, 14, 2, (2025), 104-112. DOI: https://doi.org/10.3745/TKIPS.2025.14.2.104.

Forms

Search
(IN TITLE, AUTHOR, ABSTRACT,KEYWORDS)

Advanced Search ▼

POPULAR KEYWORDS
(TOP 10 KEYWORDS)

Recent Publications
(LAST 3 YEARS)

Old Journals

Indexing

All publications of TKIPS is indexed in DOI, EBSCO, Google Scholar, Crossref, and CrossCheck.

TKIPS is also selected as the Journal for Accreditation by NRF (National Research Foundation of Korea).

Related Journals