title	emoji	colorFrom	colorTo	sdk	app_file	pinned	short_description	license
Rice Classifier ResNet50	🌾	green	yellow	streamlit	src/streamlit_app.py	false	CNN-based rice image classifier with ResNet50.	mit

🌾 Rice Classification with ResNet50 (Transfer Learning)

🔗 Live Demo: [(https://huggingface.co/spaces/EnYa32/Rice_Classifier_ResNet50)]
This app uses a deep learning model to classify rice grain images into different rice types.

The model is based on ResNet50 (Transfer Learning) and reaches very high accuracy.

---

🚀 Quick Facts

• Task: Image Classification
• Model: ResNet50 (Transfer Learning)
• Framework: TensorFlow / Keras
• Classes: 5
• Deployment: Streamlit + HuggingFace Space

---

📸 Model Insights & Training Visualizations

🔥 Grad-CAM Explainability

Model attention heatmap showing which regions influenced the prediction.

---

📈 Training Curves (Accuracy & Loss)

Training vs validation accuracy and loss across epochs.

---

🚀 Features

1ee0bca (Add model screenshots to README)

• Upload rice grain images (JPG / PNG)
• Automatic rice type classification
• Confidence score output
• Transfer learning with ResNet50 backbone
• Production-ready Streamlit interface

---

📊 Model Performance

• Validation Accuracy: 99.29%
• Top-3 Accuracy: 100%
• Low confusion between most classes
• Minor confusion between visually similar rice types

---

🌾 Rice Classes

The model can classify the following rice types:

• Arborio
• Basmati
• Ipsala
• Jasmine
• Karacadag

---

🖼️ How to use the app

1. Upload a rice image (JPG or PNG)
2. The model analyzes the image
3. The predicted rice type is shown
4. You also see the confidence score

---

🧪 Training Summary

• Base Model: ResNet50
• Transfer Learning: Yes
• Data Augmentation: Yes
• Optimizer: Adam
• Loss Function: Sparse Categorical Crossentropy
• Batch Size: 16

---

📁 Files in this Space

• app.py → Streamlit App
• best_rice_resnet50.keras → Trained Model
• README.md → This file

---

📂 Dataset

• Rice grain image dataset
• 5 rice classes
• Balanced class distribution
• Preprocessing: resize + normalization
• Augmentation: rotation, flip, zoom

---

🎯 Use Case

This project demonstrates a production-style computer vision pipeline using transfer learning with ResNet50, including training, evaluation, and deployment as an interactive Streamlit application.

---

👤 Author

Name: Enes Yamac
Date: 05.12.2025
Project Type: Deep Learning – CNN Image Classification with ResNet50

---

✅ Enjoy testing the rice classifier!
Upload an image and let the AI predict the rice type 🌾🤖