This is the web demo for the 128x128 version of the neural network for age and gender recognition. It is optimized for speed and is the best choice for use in desktop applications or edge devices. For applications on the cloud it is better to use the 224x224 version.
Age range: 0 - 100 (in one-year intervals)
Genders: male, female
Minimum face size: 30x30 pixels
Gender Recognition Accuracy: 98.1%
Age Estimation Mean Absolute Error: 5.8 years
Speed: 35 milliseconds on Intel Core i5 CPU
128x128 Model Benchmark
Gender prediction accuracy (%)
Age estimation mean absolute error (in years)
We provide free REST API for testing. If you want to evaluate our classifier, we will give you access to the server and the API protocol.
The age&gender classifier can be used with OpenCV library on both Windows and Linux.