Overview

Clarifai’s face age detection model is designed to accurately predict the age of an individual based on their facial features. This model utilizes cutting-edge machine learning techniques, the model is trained on a diverse dataset of facial images, allowing it to accurately detect age ranges from infancy to the elderly. 

Model 

The Age Demographic Model is a visual classifier model trained on cropped face images. The model uses the Clarifai ResNext model as the backbone for feature extraction and then fine-tunes the fully-connected layers to be able to predict the age ranges in the images. The model is designed to predict age ranges from 0-2, 3-9, 10-18, 19-29, 30-39, 40-49, 50-59, and 60-69. The model is trained on the 146,78 images for 60 epochs with a batch size of 4.

Dataset

The Age Demographic Model was using the Fairface dataset, which includes a diverse range of ages, races, and genders. The dataset was split into training, validation, and test sets, with 80% of the data used for training, 10% for validation, and 10% for testing. The dataset was curated by Clarifai's data strategy team and labeled using the following age ranges: 0-2, 3-9, 10-18, 19-29, 30-39, 40-49, 50-59, 60-69 and no-face. The dataset includes images from a variety of sources, including social media, stock photos, and public domain images.

Evaluation

The Age Demographic Model achieved different levels of performance in predicting age for different age groups. Here are the precision, recall, and F1 score results for each age group: 

• - 0-2 age group: Precision of 0.90, recall of 0.83, and F1 score of 0.86 
• - 3-9 age group: Precision of 0.67, recall of 0.90, and F1 score of 0.77
• - 10-18 age group: Precision of 0.88, recall of 0.84, and F1 score of 0.86 
• - 19-29 age group: Precision of 0.70, recall of 0.69, and F1 score of 0.70 
• - 30-39 age group: Precision of 0.44, recall of 0.46, and F1 score of 0.45 
• - 40-49 age group: Precision of 0.41, recall of 0.41, and F1 score of 0.41 
• - 50-59 age group: Precision of 0.47, recall of 0.41, and F1 score of 0.44
• - 60-69 age group: Precision of 0.55, recall of 0.43, and F1 score of 0.48 

Overall, the model achieved the highest precision, recall, and F1 score for the 0-2 age group, followed by the 10-18 age group. The model had slightly lower performance in predicting the 3-9 age group, with a lower precision and an F1 score. The model had the lowest performance in predicting the 30-39, 40-49, 50-59, and 60-69 age groups, with lower precision, recall, and F1 scores.

Limitation

The Age Demographic Model has some limitations, including: 

• - Limited accuracy for certain age ranges: The model may have limited accuracy for certain age ranges, such as infants and elderly individuals. 
• - Limited accuracy for certain races and genders: The model may have limited accuracy for certain races and genders, as the dataset may not be representative of all populations. 
• - Limited accuracy for certain facial expressions: The model may have limited accuracy for certain facial expressions, such as smiling or frowning.