Skip to main content

Manet: Improving Video Denoising With A Multi-Alignment Network

Yaping Zhao, Haitian Zheng, Zhongrui Wang, Jiebo Luo, Edmund Lam

  • SPS
    Members: Free
    IEEE Members: $11.00
    Non-members: $15.00
    Length: 00:11:29
03 Oct 2022

Facial expression is an essential factor in conveying human emotional states and intentions. A common strategy used for facial expression recognition (FER) is encoding expression representations from facial images. Although remarkable advancement has been made, challenges due to large variations of expression patterns and unavoidable hard samples still remain. in this paper, we propose dual-level representation enhancements (DLRE) addressing these issues. On one hand, mid-level representation enhancement (MRE) is introduced to avoid expression representation learning being dominated by a limited number of highly discriminative patterns. On the other hand, high-level representation enhancement (HRE) is introduced to alleviate the disturbance of misclassified representations especially for hard samples. The proposed method not only has stronger generalization capability to handle different variations of expression patterns but also greater discriminative power to capture the subtle distinctions of hard samples. Experimental evaluation on four popular databases, CK+, Oulu-CASIA, RAF-DB, and AffectNet, shows that our method achieves more competitive results than other state-of-the-art methods.

Value-Added Bundle(s) Including this Product

More Like This

  • SPS
    Members: Free
    IEEE Members: $11.00
    Non-members: $15.00
  • SPS
    Members: Free
    IEEE Members: $11.00
    Non-members: $15.00