AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Towards Backward-Compatible Representation Learning

Yantao Shen, Yuanjun Xiong, Wei Xia, Stefano Soatto

CVPR 2020 · Label-efficient and robust learning › Recognition and long-tail · approach: Margin-based face recognition losses · 75 citations (OpenAlex)

Check your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add the old classifier's influence loss during trainingNew face embedding model · IJB-C 1:1 verificationTAR (%)@FAR=$10^{-4}$: 77.86 -> 80.25Improved
Add the old classifier's influence loss during trainingNew face embedding model · IJB-C 1:N searchTPIR (%)@FPIR=$10^{-2}$: 59.34 -> 67.23Improved
Replace basic BCT's old-data influence loss with knowledge distillation on new classes (ablation)BCT · IJB-C 1:N searchTPIR (%)@FPIR=$10^{-2}$: 67.23 -> 69.02Improved
Use synthesized classifier weights for new classes in the influence loss (ablation)BCT · IJB-C 1:N searchTPIR (%)@FPIR=$10^{-2}$: 67.23 -> 70.70Improved
Replace the old-data influence-loss setting with knowledge distillation for new classes (ablation)BCT · IJB-C 1:1 verificationTAR (%)@FAR=$10^{-4}$: 80.25 -> 80.34Improved
Use synthesized classifier weights for new classes in the influence loss (ablation)BCT · IJB-C 1:1 verificationTAR (%)@FAR=$10^{-4}$: 80.25 -> 80.59Improved
Replace BCT with an L2 distance penalty between new and old embeddings (ablation)New face embedding model · IJB-C 1:N searchTPIR (%)@FPIR=$10^{-2}$: 67.23 -> 0.50Hurt
Train with Learning without Forgetting instead of BCT (ablation)New face embedding model · IJB-C 1:N searchTPIR (%)@FPIR=$10^{-2}$: 67.23 -> 59.27Hurt
Use ResNet-152 instead of ResNet-101 with BCT (ablation)BCT · IJB-C 1:N retrievalTPIR (%)@FPIR=$10^{-2}$: 59.34 -> 68.71Improved
Increase feature dimension to 256 while using ResNet-152 and BCT (ablation)BCT · IJB-C 1:N retrievalTPIR (%)@FPIR=$10^{-2}$: 68.71 -> 69.45Improved
Add ReLU after the new embedding (ablation)BCT · IJB-C 1:N retrievalTPIR (%)@FPIR=$10^{-2}$: 68.71 -> 17.73Hurt
Use Norm-Softmax for the old model and Cosine Margin for the new model (ablation)BCT · IJB-C 1:N retrievalTPIR (%)@FPIR=$10^{-2}$: 64.32 -> 71.16Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy