- Metric learning
- Speaker recognition
- Equal error rate metric
- Cosine loss, Contrastive loss and triplet loss
- Angular softmax
- Additive Angular Margin Loss
- Advanced triplet sampling techinques
TODO: add seminar notebook here
All links are provided on the last slide of the lecture
- Speaker Recognition Based on Deep Learning: An Overview
- FaceNet: A Unified Embedding for Face Recognition and Clustering
- Deep Metric Learning: A Survey
- VoxCeleb2: Deep Speaker Recognition
- SphereFace: Deep Hypersphere Embedding for Face Recognition
- Angular Softmax Loss for End-to-end Speaker Verification
- ArcFace: Additive Angular Margin Loss for Deep Face Recognition
- Sampling Matters in Deep Embedding Learning
- Speaker Recognition Based on Deep Learning: An Overview