EnYa32's picture
Update README.md
4a902e2 verified
|
Raw History Blame Contribute Delete
1.51 kB
---
title: DateFruitClassifier_CNN
emoji: 🌴
colorFrom: gray
colorTo: green
sdk: streamlit
app_file: src/streamlit_app.py
pinned: false
short_description: Date fruit image classifier app using CNN.
license: mit
---
# CNN Date Fruit Classification Project
## πŸ“Œ Project Overview
This project aims to classify different types of date fruits using a custom Convolutional Neural Network (CNN) trained from scratch.
## 🧠 Model Summary
- Input Size: 224x224x3 (RGB)
- Architecture: 5 Conv2D layers + MaxPooling
- Dense: 256 neurons + Dropout (0.5)
- Optimizer: Adam (lr=0.001)
- Loss: Sparse Categorical Crossentropy
- Callbacks: EarlyStopping, ReduceLROnPlateau, ModelCheckpoint
## πŸ“Š Final Results
| Metric | Score |
|--------|--------|
| **Validation Accuracy** | 85.71% |
| **Validation Loss** | 0.4426 |
## πŸš€ Conclusion
The model was successfully trained without transfer learning and shows stable performance. Future improvements can be made using augmentation or transfer learning.
## πŸ“¦ Files Included
- `best_date_cnn.keras` – Best trained model file
- `CNN_Report_updated.pdf` – Technical report
- `history.pkl` – Training history
## πŸ“š How to Run
```python
model = tf.keras.models.load_model("best_date_cnn.keras")
```
## πŸ›  Requirements
```
tensorflow
matplotlib
seaborn
numpy
scikit-learn
```
---
πŸ“„ **Download Technical Report (PDF)**
[CNN_Date_Report.pdf](./CNN_Date_Report.pdf)
---
## πŸ‘€ Author
**Enes Yamac**
Deep Learning Project
πŸ“… November 2025