EmanHussein/ResNet50_Garbage_Classifier

1# ♻️ Waste Category Image Classifier (Streamlit + Keras)2 3An interactive web app for classifying waste images into **10 distinct categories** using a custom-trained Convolutional Neural Network (CNN). Built with **TensorFlow/Keras** and deployed via **Streamlit**, this app demonstrates the power of deep learning in sustainability-oriented image classification.4 5---6 7## 🚀 Demo8 9Upload an image and instantly get a class prediction with confidence and full class-wise probabilities.10 11![demo-screenshot](https://github.com/user-attachments/assets/b10da05a-33e3-46b1-9468-bc0b88823414)12 13---14 15## 🧠 Model Summary16 17- **Architecture**: Custom CNN18- **Input Shape**: 224×224 RGB19- **Layers**:20  - Convolutional layers with BatchNorm and ReLU21  - MaxPooling22  - Global Average Pooling23  - Dense layers with Dropout24- **Optimizer**: Adam (`lr=1e-5`)25- **Loss**: Categorical Crossentropy26- **Final Layer**: Softmax (10 classes)27 28---29 30## 📂 Dataset31 32The dataset consists of images from 10 waste-related categories:33 34| Class         | Count |35|---------------|-------|36| battery       | 944   |37| biological    | 997   |38| cardboard     | 1825  |39| clothes       | 5327  |40| glass         | 3061  |41| metal         | 1020  |42| paper         | 1680  |43| plastic       | 1984  |44| shoes         | 1977  |45| trash         | 947   |46 47### 🔀 Data Split48 49| Split     | Images |50|-----------|--------|51| Training  | 15,806 |52| Validation| 1,976  |53| Testing   | 1,980  |54 55### 🧼 Preprocessing56 57- All images resized to `224×224`58- Normalized using: `rescale=1./255`59 60---61 62## 🏗️ Model Architecture (Keras Sequential)63 64```python65Input(shape=(224, 224, 3)) ➜66Conv2D(32) ➜ BN ➜ ReLU ➜67Conv2D(64) ➜ BN ➜ ReLU ➜ MaxPool(4×4) ➜68Conv2D(128) ➜ BN ➜ ReLU ➜69Conv2D(128) ➜ BN ➜ ReLU ➜ MaxPool(2×2) ➜70Conv2D(256) ➜ BN ➜ ReLU ➜ MaxPool(2×2) ➜71Conv2D(512) ➜ BN ➜ ReLU ➜ MaxPool(2×2) ➜72GlobalAveragePooling ➜73Dense(256) ➜ Dropout(0.4) ➜74Dense(10, activation='softmax')75````76 77---78 79## 🖥️ Web App Features80 81* ✅ Upload image and classify in real-time82* 📈 Display confidence score83* 📊 Class probabilities table/bar chart84* 🧠 Show model architecture summary in sidebar85 86---87 88## 📦 Installation & Run Locally89 901. **Clone the repo**91 92```bash93git clone https://github.com/Amr1Moussa/NTI-final_projects.git94cd NTI-final_projects95```96 972. **(Optional)** Create a virtual environment98 99```bash100python -m venv venv101source venv/bin/activate  # On Windows: venv\Scripts\activate102```103 1043. **Install dependencies**105 106```bash107pip install -r requirements.txt108```109 1104. **Run the app**111 112```bash113streamlit run app.py114```115 116---117 118## 📄 Requirements119 120See `requirements.txt`:121 122```txt123streamlit124tensorflow125numpy126pillow127pandas128matplotlib129```130 131## 📃 License132 133MIT License — use freely, contribute openly, credit kindly.134