A New Age Recognition API detects the age of people on photos
Everypixel Group launched the new API called Age Recognition. It’s created for those who work with the photo. We’re certainly sure that every photographer, which has ever dealt with photo stock sites, is extremely annoyed to attribute his or her content. A thousand photographers spend an enormous amount of time to describe their photo and to find the most correct keywords.
We’ve already learned our neural networks to estimate the attractiveness of Instagram bloggers’ or professionals’ photos and to generate keywords for a photo. Now we want more: let AI predict the age of people on images!
Who can use it?
It can be helpful for photographers and users of photo stock sites: Age Recognition makes keywording more accurate and image searching more relevant. The model becomes the addition to Image Keywording and Stock Photo Scoring services by Everypixel.
For example, while attributing this photo, AI suggested the next keywords:
Senior Adult
People
Women
Togetherness
Smiling
Outdoors
Retirement
Love
Adult
Eat Pray Love
Men
Caucasian Ethnicity
Lifestyles
Happiness
Cheerful
Park Ji-Sung
Gray Hair
Embracing
Females
Old
Senior Men
Nature
Portrait
Joy
Ageing Process
Two People
Enjoyment
Romance
Relaxation
Couple — Relationship
White Hair
So, you could notice that we marked in bold 8 words, which is related to an age of people. Now let’s use the Age Recognition model. And what did we see? There were 2 people: 69 and 60 years old.
If you need to submit this photo on a photo stock site, you should add those new keywords (69-year-old, 60-year-old). Bingo! Your content will be more relevant for users who find images with “you-name-it”-year-old people.
In the future, the Age Recognition will be integrated into DAM-service, which combines Image Keywording and Stock Photo Scoring services with sales analysis. The integration will make DAM more efficient and accurate in auto keywording and searching. DAM-service is based on AI, which generates non-obvious and abstract words as well as recognizes an emotional context of photos and videos and rates tags by relevance. Also, the service provides uploading and submission content to 16 stock sites and lets photographers and videographers to organize the process on the team up to 3 members and to categorize content into the projects.
Only photographers may use it?
No! We’re so excited about transforming age detection algorithms from an entertaining app to a tool for managing and improving business processes. Photographers already have taken advantage of using AI in their work. Now, this technology will continue extending in other spheres. We predict that the Age Recognition model may be used for categorization data in companies collecting information about people (such as model or recruiting agencies), for forming a selection of respondents during social science researches, for verification users of dating services, social media, and other services which require confirmation of age.
How does it work?
AI is trained on a few datasets up to 300 000 images of people of different gender, ages, and nationalities. Data scientists worked with photos of celebrities from different periods of their life. AI was taught to monitor how an appearance changes with age and predicted how older people are. There may be differences between predicted and biological age within no more than 3 years.
So,
