Capsule Networks

Why It Was Important

Proposed by Geoffrey Hinton to address CNN vulnerabilities (like failing to recognize faces if the eyes and mouth are swapped). Capsules bundle neurons together to output a vector, preserving viewpoint and orientation data mathematically. Though computationally heavy, they represent a profound alternative to standard CNNs.

Who Invented It

Geoffrey Hinton, Sara Sabour

Pioneers attempting to fix the structural flaws of CNNs.

Applications

• Pose Estimation
• Robust Image Recognition