From Keypoints to Object Landmarks via Self-Training Correspondence: A novel approach to Unsupervised Landmark Discovery

Computer Science

• New-State 100%
• Image Synthesis 100%
• Contrastive Learning 100%
• Iterative Algorithm 100%
• Unsupervised Learning 100%
• Distinctive Feature 100%
• Clustering Feature 100%

Keyphrases

• Landmark Detector 100%
• Self-training 100%
• Landmark Features 66%
• Iterative Algorithm 33%
• Contrastive Learning 33%
• Pseudo Label 33%
• Human3.6M 33%
• Feature Learning 33%
• Equivariance 33%
• Viewpoint Change 33%
• Novel Paradigm 33%
• Unsupervised Learning 33%
• Auxiliary Task 33%
• Image Generation 33%
• Distinctive Features 33%
• Feature Clustering 33%