Modification of LeNet 5 Architecture For Effective Plant Disease Detection

Simple item page
dc.contributor.advisorAsst. Prof. Sarbin Sayami; Supervisor
dc.contributor.authorMadhusudan Adhikari
dc.date.accessioned2024-08-18T04:24:54Z
dc.date.available2024-08-18T04:24:54Z
dc.date.issued2024
dc.description.abstractConvolution Neural Networks have shown great promise in computer vision. Plant disease detection through the analysis of leaf images is one of the practical applications of computer vision. To utilize such capability of CNNs effectively, research on lightweight models is necessary which makes it suitable for real-time and IoT hardware. Such hardware has low computing resources in comparison. LeNet-5 is one of the well-known pioneer CNN models and models based on its modifications have the potential for practical use in multiple areas including plant disease detection. In this work, taking an empirical approach, experimenting around multiple hyper-parameters and the structure of the classic LeNet-5, a modified model was obtained which had higher accuracy and lower training time compared to the original LeNet-5. Experiments included structural changes like the addition of batch normalization layers, additional fully connected layers, changes in the pooling mechanism and hyper-parameter changes like using ReLU and PReLU activation functions, and changes in several filters. The modified model was able to obtain a testing accuracy of 94.92% in comparison to the testing accuracy of 91.18% shown by the original LeNet-5 on the Plant Village dataset. Also, the model achieved that accuracy in almost half the time needed to train the original LeNet-5.
dc.identifier.urihttps://hdl.handle.net/20.500.14540/22710
dc.language.isoen_US
dc.subjectConvolutional Neural Networks
dc.subjectPlant Disease Detection
dc.subjectLeNet-5
dc.subjectImproving CNNs
dc.subjectHyperparameters Optimization
dc.subjectPlant Village Dataset
dc.subjectImage classification. LeNet 5 Modification
dc.titleModification of LeNet 5 Architecture For Effective Plant Disease Detection
dc.typeThesis
local.academic.levelMasters
local.institute.titleInstitute of Science & Technology
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
Modification of LeNet 5 architecture for effective plant disease detection_Madhusudan_Adhikari_CSIT (2) (1).pdf
Size:
1.43 MB
Format:
Adobe Portable Document Format
Download
License bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
license.txt
Size:
1.71 KB
Format:
Item-specific license agreed upon to submission
Description:
Download
Collections
Computer Science & Information Technology