UTKFace Age & Gender Prediction

This project builds and trains a deep learning model to predict age and gender from face images using the UTKFace dataset.
The pipeline includes dataset handling, preprocessing, custom data generation, model building (using transfer learning), training, and evaluation.

📂 Dataset

The project uses the UTKFace dataset, which contains over 20,000 labeled face images.

Each filename encodes the following information: age_gender_race_date.jpg

This project extracts only:

• Age — integer
• Gender — 0 = male, 1 = female

The dataset is downloaded automatically from Kaggle.

🧹 Preprocessing

Steps:

• Extract image files from the Kaggle ZIP archive.
• Parse labels from filenames.
• Resize images to 128×128.
• Normalize pixel values to [0, 1].
• Use a custom Keras Sequence to efficiently load data during training.

A train-validation split is performed using train_test_split.

🧱 Custom Data Generator

A subclass of keras.utils.Sequence is used to:

• Load images batch-by-batch.
• Apply preprocessing.
• Shuffle training data each epoch.
• Output:
  • X = preprocessed face images
  • y_age = age (regression)
  • y_gender = gender (binary classification)

🧠 Model Architecture

The model uses MobileNetV2 (pretrained on ImageNet) as the base feature extractor.

The top layers include:

• GlobalAveragePooling
• Dropout
• Dense layers
• Two output heads:
  • Age output → regression layer
  • Gender output → sigmoid classifier

The model is compiled with:

• mse loss for age
• binary_crossentropy for gender
• Adam optimizer

---

🏋️ Training

The model is trained with:

• Batch size: 32
• Image size: 128×128
• Validation split defined earlier

During training:

• Losses for both outputs are tracked
• Combined loss is optimized

🧪 Evaluation

After training, the model is evaluated on the validation dataset for:

• Age prediction error (MSE)
• Gender classification accuracy

▶️ How to Run

1. Install dependencies: pip install tensorflow numpy opencv-python scikit-learn kaggle
2. Place your kaggle.json API key in the working directory.
3. Run the notebook cells in order.