This project implements a Convolutional Neural Network (CNN) model to classify dental images into multiple categories using TensorFlow and Keras. The model is trained with data augmentation and evaluated on separate validation and test datasets.
📌 Project Overview
The goal of this project is to build a deep learning model capable of classifying teeth images into different categories using a custom CNN architecture.
The workflow includes:
Data preprocessing & augmentation
Loading dataset using ImageDataGenerator
Visualizing class distribution
Displaying original & augmented images
Building a CNN model
Training & validation
Testing & evaluation
Saving & loading the trained model
📂 Dataset Structure
The dataset is organized in the following directory structure:
Teeth_DataSet/ │ └── Teeth_Dataset/ ├── Training/ ├── Validation/ └── Testing/
Each folder contains subfolders representing the class labels.
Unwanted folders (e.g., out, output, outputs) are excluded during loading.
⚙️ Data Preprocessing
Images resized to: 128x128
Batch size: 32
Pixel normalization: rescale=1.0/255
Data augmentation techniques:
Rotation
Width & height shifting
Shearing
Zooming
Horizontal flipping
This improves generalization and reduces overfitting.
📊 Data Visualization
The project includes:
📈 Class distribution plots using seaborn
🖼 Display of original training images
🔄 Display of augmented images
This helps in understanding dataset balance and augmentation impact.
🧠 Model Architecture
The CNN model consists of:
3 Convolutional layers:
Conv2D (32 filters)
Conv2D (64 filters)
Conv2D (128 filters)
MaxPooling after each convolution
Flatten layer
Dense layer (128 neurons)
Dropout (0.5) to reduce overfitting
Output layer with softmax activation
Loss Function:
categorical_crossentropy
Optimizer:
Adam
Metric:
Accuracy
🚀 Training
Trained for 100 epochs
Validation dataset used during training
Accuracy and loss plotted for:
Training set
Validation set
🧪 Model Evaluation
The trained model is evaluated on a separate test dataset:
Test Accuracy: XX.XX
💾 Model Saving & Loading
The trained model is saved as:
teeth_classification_model.h5
It can be reloaded using:
load_model("teeth_classification_model.h5")
🛠 Technologies Used
Python
TensorFlow / Keras
NumPy
Matplotlib
Seaborn
OpenCV
📌 Future Improvements
Use Transfer Learning (e.g., MobileNet, ResNet)
Add EarlyStopping & ModelCheckpoint
Hyperparameter tuning
Confusion matrix & classification report
Convert model to TensorFlow Lite for deployment