Herb leaves pattern recognition using digital microscope and deep learning
Herb leaves are not fully exploited for their benefits due to the problem arise from difficulties in identifying the type of plants. Visual-based pattern recognition via microscopic imaging provides high precision and accuracy for identification that easily outperforms the limitations of human perce...
| Main Authors: | , , |
|---|---|
| Format: | Conference or Workshop Item |
| Language: | English |
| Published: |
Persatuan Saintis Muslim Malaysia (PERINTIS)
2019
|
| Subjects: | |
| Online Access: | http://irep.iium.edu.my/78345/ http://irep.iium.edu.my/78345/ http://irep.iium.edu.my/78345/13/78345%20HERB%20LEAVES%20PATTERN.pdf |
| Summary: | Herb leaves are not fully exploited for their benefits due to the problem arise from difficulties in identifying the type of plants. Visual-based pattern recognition via microscopic imaging provides high precision and accuracy for identification that easily outperforms the limitations of human perception. A three-class dataset of 255
leaf vein images - betel, noni and ortosiphon stamineus, were captured by a digital microscope was constructed as a training set for a visual-based pattern recognition
system. A simple visual-based pattern recognition system was constructed by using MATLAB and deep learning. Transfer learning was applied on a pretrained VGG-16
network with a validation accuracy of 91.03%. Further fine-tuning was applied on the network resulting an increase of accuracy to 98.72%. |
|---|