Age-Gender Detector

• Age-Gender Detector
  • Data
    • Data Explanation
  • Model Full
  • Model Cropped
    • Dash images
  • Steps to run web app

Data

UTKFace dataset

UTKFace dataset Kaggle

This dataset contains over 20,000 face images with annotations of age, gender, and ethnicity. The images cover large variation in pose, facial expression, illumination, occlusion, resolution, etc.

Data Explanation

If a photo is called 34_0_0_201701171712010149082.jpg.chip.jpg, it means that the age of the individual is 34 and his gender is male. That is, the photo names follow the following scheme age_gender_race_relevant_data.jpg.chip.jpg.

Gender being 0 for male and 1 for female.

Model Full

We have used Convolutional Neural Networks (CNN) to predict the age and gender of the full input image.

We have obtained the following metrics:

• Validation:
  • Gender accuracy: 0.869
  • Age MSE: 135.65
  • Age MAE: 8.54
• Test:
  • Gender accuracy: 0.878
  • Age MSE: 140.35
  • Age MAE: 8.68

Model Cropped

We have used Convolutional Neural Networks (CNN) to predict the age and gender of the cropped input image.

We have obtained the following metrics:

• Validation:
  • Gender accuracy: 0.905
  • Age MSE: 78.956
• Test:
  • Gender accuracy: 0.915
  • Age MSE: 75.28

Dash images

Prediction tool (the age has an error margin):

Saliency maps:

Model metrics (best model selected):

Steps to run web app

From the main project folder (AgeGenderDetector/) run the following command:

pip install -r requirements.txt

Then, from one of the main folders (AgeGenderDetector/full/ or AgeGenderDetector/cropped/) run the following command:

python -m dashboard.dashboard