Mobile-Based Inventory Counting Using 2D Object Detection and Layer-Based Inference
Pertanika Journal of Science and Technology · 30.10.2026 · 34 (5)
DOI: 10.47836/pjst.34.5.23 ↗Kurz gefasst
Fachartikel in der Zeitschrift Pertanika Journal of Science and Technology, online erschienen am 30.10.2026. Verfasst von Heng Lai Gan und 2 weiteren Autor:innen.
Abstract (Originalsprache)
The term “2D” refers to layer-based quantity inference from a spatially stacked arrangement and does not indicate direct 2D bounding-box regression or depth reconstruction. YOLOv8 is a deep learning framework that enables the development of exceedingly fast computer vision applications. It is adept at object detection and classification, including multi-object detection. Computer Vision explores methods for reconstructing, interpreting, and understanding a scene's three-dimensional structure from its two-dimensional images, leveraging the scene's inherent structure. It involves modelling and simulating human vision using both hardware and software. In this project, YOLOv8 was used as the processing algorithm, integrated with other tools and libraries to build a mobile-based inventory-counting application using 2D object detection and layer-based inference. It is a design that will be used for stock checking or dispatching goods from stock. The input will be the mobile phone's front camera for capturing the boxes. The scene will be captured from the surrounding area of the stock area, where the rack is located. To train the model, a custom dataset of 1,500 images containing three cardboard box sizes was utilised, annotated via the Roboflow platform. Separate YOLOv8 models were trained for the Small, Middle, and Big box datasets. Bounding-box detection was evaluated on labelled validation images using mAP@0.5 and mAP@0.5:0.95, while the complete mobile inventory-counting application was evaluated separately through 250 controlled trials for each box-size model. Finally, the system demonstrated its effectiveness in application. It accurately recognised and measured various box types, assuming there were no vacant spaces within a 3 × 3 × 3 stacking configuration. The variation in box size enables an in-depth evaluation of the device’s effectiveness across a range of in-stock items.
Zitieren
APA
Gan, H. L., Goh, E. S., & Ahmad, J. (2026). Mobile-Based Inventory Counting Using 2D Object Detection and Layer-Based Inference. Pertanika Journal of Science and Technology, 34(5). https://doi.org/10.47836/pjst.34.5.23
Vancouver
Gan HL, Goh ES, Ahmad J. Mobile-Based Inventory Counting Using 2D Object Detection and Layer-Based Inference. Pertanika Journal of Science and Technology. 2026;34(5). doi:10.47836/pjst.34.5.23
BibTeX
@article{Gan2026,
author = {Gan, Heng Lai and Goh, Eg Su and Ahmad, Johanna},
title = {Mobile-Based Inventory Counting Using 2D Object Detection and Layer-Based Inference},
journal = {Pertanika Journal of Science and Technology},
year = {2026},
volume = {34},
number = {5},
doi = {10.47836/pjst.34.5.23}
}
EUVIMED