Improving Noisy Face Embeddings Using Time Series Transformers and Adaptive Triplet Loss
Face recognition accuracy is now at the point that it is comparable to human performance, largely driven by improvements in face embedding models. However, face embedding models rely on high-quality images, making them vulnerable to variations in quality, illumination, pose, and occlusion in real-wo...
Kaydedildi:
| Asıl Yazarlar: | , , |
|---|---|
| Materyal Türü: | Artigo |
| Dil: | Inglês |
| Baskı/Yayın Bilgisi: |
IEEE
2025-01-01
|
| Seri Bilgileri: | IEEE Access |
| Konular: | |
| Online Erişim: | https://ieeexplore.ieee.org/document/11193818/ |
| Etiketler: |
Etiket eklenmemiş, İlk siz ekleyin!
|