Everypixel, the company behind various AI-based tools for image analysis and processing –that is also under the umbrella of the Pressfoto Group–, has a new addition to their API service for stock photo providers: a software that recognizes the age of people in photos.
Useful to both photo buyers and photographers, this scientifically-trained tool will later be available in Everypixel’s DAM product for portfolio management too.
Here we have some more info on this cool new feature and how it works!
Everypixel Age Recognition: Detecting Age in Faces Automatically
Everypixel is Pressfoto Group’s intelligent image recognition lab, dedicated to developing software tools that serve to save time and effort when working with big photo collections. For example, one of their first products was the Everypixel cross-agency search engine that lets you find images from multiple image banks at once and compare prices.
The new Age Recognition feature is part of Everypixel’s API solution for image providers (such as stock photo agencies), that includes automatic keywording and photo scoring for both professional and UGC images. The new tool uses intelligent machine learning to identify the traits in human faces and match them to a certain age.
In the API demo model, you can test the function by simply uploading or dragging-and-dropping any image you like (they provide you with a few sample photos too) and selecting the “age recognition” tab. The software takes seconds to analyse the image and display a number that represents the age of the person/people portrayed.
We’ve put it to test and in our experience the results, while not 100% accurate, are within a very reasonable age range.
Want to see it in action? Test Everypixel’s Age Recognition Model Demo right here.
AI Trained to Know How Human Age Manifests in Faces
According to the Everypixel team, their data scientists have trained the new feature by letting it analyse a database of around 300,000 images sourced from IMBD and Wikipedia, two sources that display various portraits of celebrities from year to year, showing their age progression.
This is how the system monitors the changes in human faces at different life stages and learns to identify the markers that signal each age. Very cool.
We think this data training explains the slight imprecision of the results since celebrities tend to look younger or older in photographs according to styling, makeup, lighting, retouching, etc.
Still, the ages recognised are fairly accurate and useful for the purposes of image search or categorization. The best stock photo agencies already offer an age filter for their search engines, and this tool could make that functionality more intuitive and easier for more companies to offer.
Plus, let’s be honest, it’s also fun to use!
Give it a go yourself and try the Everypixel Age Recognition Demo with your favourite photos.
