Citation

Mohamed Ahmed Mansour, Mohamed yasser and Dambul, Katrina D. and Choo, Kan Yeep (2022) Oil Palm Fresh Fruit Bunch Ripeness Classification and Localization Using Object Detection Algorithms. In: Postgraduate Colloquium December 2022, 1-15 December 2022, Multimedia University, Malaysia. (Unpublished)

Text Mohamed Yasser-foe.pdf - Submitted Version Restricted to Repository staff only Download (682kB)

Abstract

Oil palm bunches ripeness is an important factor for the production proves. Relying on human worker often results in misclassification It must be in the correct ripeness to yield the best oil quality. Current practice relies on human workers to harvest the bunches. Using computer vision can help improve the harvesting process and avoid inaccurate grading. In this study, YOLOv5 algorithms was test on the task of ripeness classification of oil palm fresh fruit bunches. Several improvement were made to YOLOv5s to improve its performance such as label smoothing, image augmentation, Squeeze and Excitation (SE), convolution block attention mechanism (CBAM) and RepVGG. All testing was trained on the same dataset of four ripeness classes: unripe, half ripe, ripe and over ripe. Models were compared by mean average precision, precision, recall, F1 Score and model parameters and parameters size. Overall, YOLOv5s with external image augmentation and SE layer has shown an improvement in mAP and precision. Future work on hyperparameters optimization will help to improve the model performance.

Item Type:	Conference or Workshop Item (Poster)
Uncontrolled Keywords:	Oil palm
Subjects:	S Agriculture > SB Plant culture
Divisions:	Faculty of Engineering (FOE)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	29 Dec 2022 01:05
Last Modified:	29 Dec 2022 01:05
URII:	http://shdl.mmu.edu.my/id/eprint/11064

Downloads

Edit (login required)