0% found this document useful (0 votes)
27 views19 pages

CSE523 Presentation

Good presentation

Uploaded by

yadullahgoni
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
27 views19 pages

CSE523 Presentation

Good presentation

Uploaded by

yadullahgoni
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 19

Rice Leaf Disease

Detection and
Classification
Presented By:
Name: Katura Gania Khushbu
ID: 2023-1-96-011
Name: Nabila Islam
ID: 2023-3-96-011
About Rice
❑ Rice is a crucial staple food, providing
nutrition for billions and supporting
economies worldwide.
❑ It holds cultural significance and is
extensively traded internationally,
ensuring food security.
❑ While versatile and nutritious, its
cultivation can have environmental
impacts.
❑ Overall, rice plays a vital role in global
food security, economics, culture, and
the environment.
Why rice leaf is important for production of rice

Worldwide Bangladesh
▪ The Crucial Role of Rice ▪ Agricultural Backbone
Leaves in Photosynthesis.
▪ Economic Impact
▪ Maximizing Crop Yields.
▪ Food Security
▪ Protecting Profitability.
▪ Cultural Significance
▪ Securing Rice Leaf Health.
▪ Export Potential
▪ Optimizing
Photosynthesis.
Why we need to
detect and classify
rice leaf disease

1. Early Detection
2. Crop Management
3. Yield Loss Prevention
4. Economic Impact
5. Food Security
6. Environmental Sustainability
7. Research and Development
Selected Research Paper
Optimization Based Rice Leaf Disease Classification in Federated
Learning

Overview: This paper proposes a Federated Learning (FL)-based classification


scheme for rice leaf diseases, addressing the significant impact of such diseases
on the rice industry's sustainability.

Dataset: The dataset used comprises 40 images in each class of rice leaves affected
by various diseases, brown spot, bacterial leaf blight, and leaf smut.

Image Processing: The input images undergo preprocessing using a Kalman filter to
remove noise, followed by data augmentation using random image warping,
cropping, and colour change. After that, CNN features, statistical features, LTP,
SLIF, SURF, and LOOP features are used to extract features.
Selected Research
Paper

Classifiers: LeNet which was trained using the suggested SHAO is used to classify diseases.
By combining SHO and AOA in this case, a novel SHAO is proposed. The extracted features
are utilized in the Proposed SHAO_Lenet for Classification of Rice Leaf Disease with FL.

Accuracy: The proposed method achieves a precision of 91.3%, recall of 92.2%, and
f-measure of 91.7%, with a loss function of 3.3%, Mean Square Error (MSE) of 7.3%, and
Root MSE of 27.1%.

Limitations: Incorporating a wider range of rice diseases in the dataset could enhance the
model's ability to generalize and classify different types of leaf abnormalities accurately.
Research
model
Methodology
.

.
Experimental results
using
a) Input image
b) Pre-processed image
c) Augmented image
produced with color
augmentation
d) Augmented image
produced with cropping
e) Augmented image
.
produced with random
image warp

.
Experimental
outcomes using

a) LOOP feature
extracted image
b) LTP feature
extracted image
c) SURF feature
extracted. image

.
Competition Analysis of
related paper
Ref No. Image Processing/ Data Methods Dataset Accuracy
Processing Techniques

[1] Aspect ratio: 9:1 Faster-RCNN 3 Classes, 1800 87.2%


Frame size: 1920 × 1080 pixels Framework, Images
Max size: 2500 pixels DCNN backbone
Min size: 1500 pixels was trained with
Detection threshold: < 0.2 CUDA 9.1 and
cuDNN 7.1.3,

[2] Resized to 224 × 224 DenseNet, 515 images 92.86 %,


Image sharpening and edge filling VGGNet, Inception affected by 13
Ratio: 7:3 V3, ResNet diseases

[3] - Resized to 227 × 227 × 3 Transfer Learning Dataset Name: 97.96%


- Augmentation: Rotations and CNN Models: On-Field Rice
flipping AlexNet, VGG16, Leaf Disease
- Random split: 80:20 VGG19, and 8 other Dataset,5932
CNN models images and 4
[4] Used 32 pre-trained models for feature Decision Trees, 551 images with 3 94%
extraction and applied diverse machine k-Nearest Neighbors, classes
learning and ensemble algorithms for Random Forest
disease classification.

[5] Hyperspectral Imaging (HSI) Deep Transfer Hyperspectral


Learning images of rice 88%
leaves from four
different varieties:
[6] Enhancement,Augmentation Transfer learning with 11,022 images with 96.08%
Background removal, resizing, VGG19 fine 6 classes
enhancement, data augmentation

[7] - Feature Extraction: Utilized 32 Decision Trees, 521 images and 3 94%
pre-trained models k-Nearest Neighbors, classes
- Segmentation: Applied on images Random Forest
- Classification: Employed diverse
machine learning and ensemble
algorithms
[8] Image augmentation, Support vector 6 classes 91%
BGR to RGB, RGB to machine classifier
Gray, and convolutional
gaussian smoothing neural networks

[9] Cropping, Random SHAO_Lenet 40 images in each precision of 91.3%,


Image Warping, and class with 4 recall of 92.2%,
Color Transformations. classes f-measure of
91.7%

[10] Splitting, Modified YOLO v8 1634 images with 89.9%


augmentation(flipping, architecture 3 classes
translation, and mosaic)
Research gaps

• More varieties of rice diseases and more fine-tuned CNN


models could be used.
• Video detection systems had a relatively long detection
time and relatively low detection confidence.
• Limited discussion on specific techniques used for
background removal, resizing, and enhancement,
requiring clarification on augmentation methods
employed.
• Lack of validation on real-world datasets or field
conditions may affect the generalizability of results to
practical agricultural settings.
• Limited dataset size of diseased samples may restrict
the abilities of machine learning and deep learning
methods.
• Specific disease classifications, such as leaf folder and
blast disease, require improved accuracy.
Dataset Overview
Ref. no. Leaf Disease No of Images
[1] [2] [3] Bacterial Leaf Blight 40 + 1584 +1197
[1] [2] [3] Brown spot 40 + 1600 +1546
[1] Leaf smut 40
Total Class=11
[1] [3] Blast 1440 +1748
Total [2] Tungro 1308
Image=17,842 [3] Hispa 1299
[3] Narrow brown spot 954
[3] Leaf Scaled 1332
[3] Neck Blast 1000
[3] Sheath Blight 1629
[3] Healthy Rice Leaf 1085
Data Analysis Plan
Proposed Methodology
Proposed Model

The planned steps for this project:

Step 1: Data Collection


Step 2: Data Preparation
Step 3 : Data Preprocessing
Step 4: Designing Model Architecture
Step 5: Models Experimentations
Step 6: Model Comparative Analysis
Conclusion

● Bangladesh heavily relies on agriculture for its economy.


● Rice Crop diseases can lead to substantial yield losses and impact the livelihood of
farmers.
● 10 research paper has been summarized from which three dataset has been selected
and merged.
● The dataset consists of 17,842 images of 11 classes (10 diseases and 1 healthy).
● The datasets contributed would be useful to students and researchers to investigate
further on development of algorithm and deep learning concepts to educate more
about Rice diseases.
Resources
[1] Li, Dengshan, et al. "A recognition method for rice plant diseases and pests video detection based on deep
convolutional neural network." Sensors 20.3 (2020): 578.
[2] Chen, Junde, et al. "Detection of rice plant diseases based on deep transfer learning." Journal of the Science of
Food and Agriculture 100.7 (2020): 3246-3256.
[3] Sethy, Prabira Kumar, et al. "Deep feature based rice leaf disease identification using support vector
machine." Computers and Electronics in Agriculture 175 (2020): 105527.
[4]. Feng, Lei, et al. "Hyperspectral imaging combined with deep transfer learning for rice disease
detection." Frontiers in Plant Science 12 (2021): 693521
[5] https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2021.693521/full
[6] https://www.mdpi.com/2223-7747/11/17/2230
[7] https://www.frontiersin.org/journals/plant science/articles/10.3389/fpls.2021.693521/full
[8] Haridasan, A., Thomas, J. & Raj, E.D. Deep learning system for paddy plant disease detection and classification.
Environ Monit Assess 195, 120 (2023).
[9]Tripathy, Ramamani, et al. "Optimization Based Rice Leaf Disease Classification in Federated Learning."
Multimedia Tools and Applications (2024): 1-27.
[10]https://www.mdpi.com/2624-7402/6/1/18
[11] https://archive.ics.uci.edu/dataset/486/rice+leaf+diseases
[12] https://data.mendeley.com/datasets/fwcj7stb8r/1
[13] https://www.kaggle.com/datasets/loki4514/rice-leaf-diseases-detection
Thank you

You might also like