File size: 3,016 Bytes
e15366b
b48bfd3
18b37d3
e15366b
 
18b37d3
 
e15366b
b48bfd3
8b56483
e15366b
 
b48bfd3
e15366b
b48bfd3
e15366b
18b37d3
 
 
 
 
b48bfd3
18b37d3
 
 
 
 
 
b48bfd3
18b37d3
b48bfd3
18b37d3
b48bfd3
18b37d3
 
 
 
 
 
 
b48bfd3
 
 
18b37d3
 
 
 
b48bfd3
18b37d3
 
 
 
 
 
 
b48bfd3
 
 
18b37d3
 
 
 
 
 
 
 
 
 
 
 
b48bfd3
18b37d3
 
 
 
 
 
 
 
b48bfd3
18b37d3
 
b48bfd3
18b37d3
b48bfd3
18b37d3
b48bfd3
18b37d3
b48bfd3
18b37d3
 
 
 
 
 
 
 
 
b48bfd3
18b37d3
 
 
 
 
 
 
 
 
 
 
b48bfd3
 
18b37d3
 
 
 
 
b48bfd3
 
18b37d3
 
 
b48bfd3
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
---
title: GrapeVineClassification_TL
emoji: πŸ‡
colorFrom: red
colorTo: red
sdk: streamlit
app_file: src/streamlit_app.py
pinned: false
short_description: Transfer Learning model for grape leaf disease.
license: mit
---

# πŸ‡ Grapevine Leaf Disease Classifier (CNN – Transfer Learning with ResNet50)

This web app demonstrates a **Convolutional Neural Network using Transfer Learning (ResNet50 pretrained on ImageNet)** to classify grapevine leaf images into one of the following four disease categories:

- **Black Rot**
- **Esca (Black Measles)**
- **Leaf Blight (Isariopsis Leaf Spot)**
- **Healthy**

The model was developed using **TensorFlow/Keras** and fine-tuned on a custom grapevine leaf dataset. It is deployed via **Streamlit on HuggingFace Spaces** and predicts disease labels with class probabilities.

---

## πŸš€ How to Use

1. Upload a grapevine leaf image (`.jpg`, `.jpeg`, `.png`)
2. Click **β€œPredict Disease”**
3. The app will display:
   - πŸ” Predicted disease  
   - πŸ“Š Confidence score  
   - πŸ“ˆ Probability distribution across all classes

---

## 🧠 Model Details

| **Attribute**       | **Value** |
|--------------------|-----------|
| Model Type         | CNN with Transfer Learning (ResNet50) |
| Pretrained Weights | ImageNet |
| Input Image Size   | 160 Γ— 160 pixels (adapted for ResNet) |
| Loss Function      | `categorical_crossentropy` |
| Optimizer          | Adam |
| Metrics            | Accuracy, Precision, Recall |
| Framework          | TensorFlow / Keras |
| Deployment         | Streamlit (HuggingFace Spaces) |
| Classes            | 4 total |

---

## πŸ“‚ File Structure

β”œβ”€β”€ src/
β”‚ └── streamlit_app.py # Streamlit web app
β”œβ”€β”€ grape_cnn_TL_ResNet50_final.keras # Final saved model
β”œβ”€β”€ grape_training_history.pkl # Training history
β”œβ”€β”€ README.md # Project documentation
└── requirements.txt # Python dependencies

yaml
Code kopieren

---

## πŸ› οΈ Technologies Used

- **Python**
- **TensorFlow / Keras**
- **ResNet50 (Transfer Learning)**
- **NumPy, PIL**
- **Streamlit**

---

## πŸ“Š Example Output

Predicted Disease: Esca (Black Measles)
Confidence: 45.25%
Probability per Class:

Black Rot β†’ 17.96%

Esca (Black Measles) β†’ 18.90%

Leaf Blight β†’ 17.89%

Healthy β†’ 45.25%

yaml
Code kopieren

---

## πŸ‘€ Author

Developed by **Enes Yamac (EnYa32)**  
as part of a Deep Learning project involving custom CNNs and Transfer Learning techniques for agricultural image classification.

---

## πŸ“Œ License

This project is released under the **MIT License**.

---

## πŸ’‘ Notes

- This version uses **Transfer Learning with ResNet50**.
- A separate version using a **pure CNN (trained from scratch)** is available as another Space.

---

## πŸ”— Related Projects

πŸ”Ή **Grapevine Disease Classifier – CNN (No Transfer Learning)**  
πŸ”Ή Future enhancements: Grad-CAM integration & model comparison

---

✨ *Thank you for testing this Transfer Learning version!*  
Your feedback is appreciated!