PROJECT 03
AI Emotion Explorer
See how an AI model classifies the visible facial expression in a photograph.
Google Gemini vision (facial expression classification)
Important: AI cannot read minds!
This AI only looks at the shape of a face in a picture (like a smile or raised eyebrows). It cannot reliably tell how someone actually feels, and it says nothing about their personality.
Lighting, camera angle, image quality and the fact that everyone shows feelings differently all change the result.
Please only use voluntary demonstration photos of people who agreed to take part. This tool does not identify people, and photos are deleted right after analysis.
1. Choose a face photo
Tip: Use a clear, front-facing photo of one face, close up and well lit. Try making different faces and compare the results!
Your result will appear here
Upload a photo and press Analyze Expression.
What is image classification?
Image classification means sorting a picture into one of several categories. This model was trained on thousands of example face pictures that people had labelled as happy, sad, angry, surprised, neutral, fearful or disgusted. It learned which patterns of pixels (like curved lips or wide eyes) often appear in each category.
For a new photo, it gives a score for every category. The scores add up to 100%. The highest score is its best guess - but a guess about a facial expression, not about real feelings. People can smile when they are nervous, or look neutral when they are happy!