/Age-and-Gender-Classification-using-Resnet50

Primary LanguageJupyter Notebook

Age And Gender Classification using ResNet-50

Description

This project utilizes a pre-trained ResNet-50 model for age and gender prediction on the UTKFace dataset. The model is fine-tuned to predict the age and gender of individuals in facial images. The age is predicted as a regression task, while gender is treated as a binary classification problem.

Requirements

Make sure you have the following dependencies installed:

  • Python (>=3.6)
  • TensorFlow (>=2.0)
  • NumPy
  • Pillow (PIL)
  • scikit-learn

Usage

  1. Dataset Preparation:

    • Download the UTKFace dataset.
    • Organize the dataset into a directory named "UTKFace."

  2. Preprocessing:

    • Adjust the batch_size variable to your liking.
    • Run the preprocessing script to load and process the images.

  3. Model Training:

    • The pre-trained ResNet-50 model is loaded and fine-tuned on the UTKFace dataset.
    • Age is predicted using Mean Squared Error loss, and gender is predicted using Categorical Crossentropy loss.

  4. Save Model:

    • The fine-tuned model is saved as 'model.h5.'

Testing

The Python script (gui.py) provides a graphical user interface (GUI) for testing the trained deep learning model used for gender and age prediction. The model has been fine-tuned on the UTKFace dataset and is loaded from the 'model.h5' file.

Requirements

Make sure you have the following dependencies installed:

  • Python (>=3.6)
  • OpenCV
  • NumPy
  • Keras
  • Pillow (PIL)
  • Matplotlib
  • Tkinter

Ensure that the 'model.h5' file containing the trained model is available.

Usage

  1. Run the GUI(gui.py file):

    • Execute the script to launch the Tkinter GUI.
    • Use the "Real-time" button to start capturing video from your webcam with real-time predictions.
    • Alternatively, use the "Upload an Image" button to select an image for prediction.

  2. Real-time Prediction:

    • Press 'q' to exit the real-time prediction loop.

  3. Image Upload:

    • Select an image using the file dialog.
    • The application will display the original image and predictions for gender and age.

Notes

  • The application utilizes a pre-trained ResNet-50 model fine-tuned for gender and age prediction.
  • Real-time prediction uses OpenCV for video capturing.
  • Tkinter is used for the graphical user interface.
  • Ensure the necessary dependencies are installed before running the application.