COMPUTER VISION / FER2013
A face.
Seven expressions.
Explore what a neural network sees in a smile, a frown, or a moment of surprise. Facial expression recognition, right in your browser.
Small details.
Different predictions.
01 / BEHIND THE PROJECT
Learning the language
of facial expressions.
This XMUM deep learning project compares a custom convolutional neural network with VGG16, ResNet50, and MobileNetV2 on FER2013. Each model predicts seven expression classes from a 48 × 48 RGB face crop.
Explore the saved evaluation plots, then try an image or your webcam. Faces are located first; the trained classifier then scores each expression.
Predictions describe visual expressions and can be wrong. They do not establish how a person feels.
02 / TRY IT YOURSELF
Your expression. In focus.
Upload a photo or turn on your camera. Images and video stay on your device.
03 / MODEL BENCHMARKS
Four approaches.
One shared dataset.
Recorded project results on FER2013, alongside the original evaluation figures.
Test accuracy
HIGHER IS BETTERInspect the evidence
04 / LET’S CONNECT
Curious about
the project?
Have a question, an idea, or an opportunity? Leave a message.
Explore the source on GitHub ↗