Plant disease detection using deep learning
The success of deep learning (DL) has greatly promoted the use of computer vision technology in smart agriculture. Many developments in this area are focusing on timely and accurate recognition of plant disease, with the goals of increasing crop productivity and fostering economic growth. This pro...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Year Project / Dissertation / Thesis |
| Published: |
2024
|
| Subjects: | |
| Online Access: | http://eprints.utar.edu.my/6553/1/MH_1902674_Final_CHENG_JUNG_YIN.pdf http://eprints.utar.edu.my/6553/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Universiti Tunku Abdul Rahman |
| Summary: | The success of deep learning (DL) has greatly promoted the use of computer vision technology in smart agriculture. Many developments in this area are focusing on timely and accurate recognition of plant disease, with the goals of
increasing crop productivity and fostering economic growth. This project aims to explore the potential of two popular DL architectures namely You Only Look Once (YOLO) and transformer for recognizing 70 distinct classes of plant leaf health conditions. Specifically, a total of six models namely Vision Transformer-B/16 (ViT-B/16), ViT-B/32, ViT-L/16, YOLOv8n-cls, YOLOv8s-cls and YOLOv8m-cls are implemented and compared. In the training stage where graphics processing unit (GPU) is utilized, ViT-B/32 yields
the shortest training time, which is at least 80% faster than all YOLOv8-cls variants. However, when deploying these trained models on a central processing unit (CPU), YOLOv8 models consistently outperform ViT algorithms in terms
of speed and accuracy. Experiment results indicate that YOLOv8n-cls attains the highest frames per second (FPS) of 31, whereas YOLOv8m-cls achieves a test accuracy of 97.7 %. Such findings suggest that YOLOv8 appears to be more
promising for real-time object classification tasks.
|
|---|