♻️ Garbage Classification – CNN Deep Learning Project

Upload a photo of waste → the CNN model classifies it and tells you how to dispose of it.

---

🔧 FIX: "Model Not Found" Error

This is the most common error when loading Keras .h5 models. Three lines of code fix it permanently:

import os

# Build the path relative to THIS file, not the working directory
BASE_DIR   = os.path.dirname(os.path.abspath(__file__))
MODEL_PATH = os.path.join(BASE_DIR, "waste_classifier_model.h5")

# Then load:
import tensorflow as tf
model = tf.keras.models.load_model(MODEL_PATH)

Why does it happen?

Cause	Broken code	Fixed code
Relative path – Python looks in CWD, not the script's folder	load_model("model.h5")	load_model(os.path.join(BASE_DIR,"model.h5"))
Wrong Keras API (TF 2.x vs standalone Keras 3)	import keras; keras.models.load_model(...)	Try tf.keras first, then keras as fallback
Model not copied to deployment folder	File missing in production	Copy .h5 next to app.py and use BASE_DIR path
TF version mismatch	Model saved with TF 2.9, loaded with TF 2.3	pip install --upgrade tensorflow

---

📁 Project Structure

garbage_classification_project/
│
├── app.py                          # Flask web app (main entry point)
├── requirements.txt
├── waste_classifier_model.h5       # ← copy YOUR model here
│
├── utils/
│   └── model_loader.py             # Robust loader utility (import anywhere)
│
├── templates/
│   └── index.html                  # Web UI
│
├── static/
│   ├── css/style.css
│   └── js/app.js
│
└── uploads/                        # Temp upload directory (auto-created)

---

🚀 Quick Start

1 – Install dependencies

pip install -r requirements.txt

2 – Place your model

# Copy your model into the project root (same folder as app.py)
cp /path/to/waste_classifier_model.h5  garbage_classification_project/

3 – Run

cd garbage_classification_project
python app.py
# → Open http://localhost:5000

4 – Verify model loads

curl http://localhost:5000/health
# → {"status":"ok","model_path":"...","model_exists":true}

---

⚙️ Configuration

Edit the top of app.py to match your model:

# Input size used during training
IMG_SIZE = (224, 224)   # change to (128,128) or (150,150) if needed

# Class labels – must match your training order
CLASS_LABELS = {
    0: {"name": "Cardboard", "emoji": "📦", "type": "Recyclable", "color": "#8B4513"},
    1: {"name": "Glass",     "emoji": "🍾", "type": "Recyclable", "color": "#00CED1"},
    # … add or remove to match your dataset classes
}

---

🧠 Model Architecture (reference)

The project expects a standard CNN .h5 saved with model.save("waste_classifier_model.h5").

Typical training pattern:

import tensorflow as tf

model = tf.keras.Sequential([
    tf.keras.layers.Conv2D(32, (3,3), activation='relu', input_shape=(224,224,3)),
    tf.keras.layers.MaxPooling2D(2,2),
    tf.keras.layers.Conv2D(64, (3,3), activation='relu'),
    tf.keras.layers.MaxPooling2D(2,2),
    tf.keras.layers.Conv2D(128,(3,3), activation='relu'),
    tf.keras.layers.MaxPooling2D(2,2),
    tf.keras.layers.Flatten(),
    tf.keras.layers.Dropout(0.5),
    tf.keras.layers.Dense(256, activation='relu'),
    tf.keras.layers.Dense(NUM_CLASSES, activation='softmax'),
])

model.compile(optimizer='adam',
              loss='categorical_crossentropy',
              metrics=['accuracy'])

# After training:
model.save("waste_classifier_model.h5")   # saves to CWD

---

🌐 API Reference

Endpoint	Method	Body	Response
/	GET	—	Web UI
/predict	POST	multipart/form-data with file field	JSON result
/predict	POST	application/json with {"image":"data:image/...;base64,..."}	JSON result
/health	GET	—	{"status":"ok","model_exists":true}

Example response:

{
  "class_index": 4,
  "class_name":  "Plastic",
  "emoji":       "🧴",
  "waste_type":  "Recyclable",
  "color":       "#FF6347",
  "confidence":  91.3,
  "disposal_tip":"♻️ Place in the blue recycling bin. Rinse containers before recycling.",
  "top3": [
    {"name":"Plastic",   "confidence":91.3},
    {"name":"Metal",     "confidence":5.1},
    {"name":"Cardboard", "confidence":2.4}
  ]
}

---

🐛 Troubleshooting

Error	Cause	Fix
OSError: No such file or directory	Relative path	Use BASE_DIR pattern (see top of README)
ModuleNotFoundError: No module named 'tensorflow'	Not installed	pip install tensorflow
ValueError: Unknown layer	Custom layers in model	Pass custom_objects={"MyLayer": MyLayer} to load_model
Low confidence on all classes	Wrong IMG_SIZE	Set IMG_SIZE to match training input shape
All predictions same class	Wrong class order	Reorder CLASS_LABELS to match model.fit generator

---

📜 License

MIT – free for personal and commercial use.