MTCNN

📖 Full definition

MTCNN (Multi-Task Cascaded Convolutional Networks) detects faces in an image and simultaneously locates landmarks (eyes, nose, mouth). It is a standard step before any embedding computation: without precise face detection, the embedding will be noisy.

💡 Concrete example

When you upload a photo to Look Like Me, MTCNN first detects the face, then aligns it, before computing its embedding.

🔄 Synonyms

Multi-Task Cascaded CNN

🔗 See also

• Facial alignment →
• Facial embedding →
• CNN (Convolutional Neural Network) →