Picture for Gilles Hénaff

Gilles Hénaff

Look Beyond Bias with Entropic Adversarial Data Augmentation

Add code
Jan 10, 2023
Viaarxiv icon

Learning Less Generalizable Patterns with an Asymmetrically Trained Double Classifier for Better Test-Time Adaptation

Add code
Oct 17, 2022
Figure 1 for Learning Less Generalizable Patterns with an Asymmetrically Trained Double Classifier for Better Test-Time Adaptation
Figure 2 for Learning Less Generalizable Patterns with an Asymmetrically Trained Double Classifier for Better Test-Time Adaptation
Figure 3 for Learning Less Generalizable Patterns with an Asymmetrically Trained Double Classifier for Better Test-Time Adaptation
Figure 4 for Learning Less Generalizable Patterns with an Asymmetrically Trained Double Classifier for Better Test-Time Adaptation
Viaarxiv icon

Test-Time Adaptation with Principal Component Analysis

Add code
Sep 13, 2022
Figure 1 for Test-Time Adaptation with Principal Component Analysis
Figure 2 for Test-Time Adaptation with Principal Component Analysis
Figure 3 for Test-Time Adaptation with Principal Component Analysis
Figure 4 for Test-Time Adaptation with Principal Component Analysis
Viaarxiv icon

Swapping Semantic Contents for Mixing Images

Add code
May 20, 2022
Figure 1 for Swapping Semantic Contents for Mixing Images
Figure 2 for Swapping Semantic Contents for Mixing Images
Figure 3 for Swapping Semantic Contents for Mixing Images
Figure 4 for Swapping Semantic Contents for Mixing Images
Viaarxiv icon

Encouraging Intra-Class Diversity Through a Reverse Contrastive Loss for Better Single-Source Domain Generalization

Add code
Jun 15, 2021
Figure 1 for Encouraging Intra-Class Diversity Through a Reverse Contrastive Loss for Better Single-Source Domain Generalization
Figure 2 for Encouraging Intra-Class Diversity Through a Reverse Contrastive Loss for Better Single-Source Domain Generalization
Figure 3 for Encouraging Intra-Class Diversity Through a Reverse Contrastive Loss for Better Single-Source Domain Generalization
Figure 4 for Encouraging Intra-Class Diversity Through a Reverse Contrastive Loss for Better Single-Source Domain Generalization
Viaarxiv icon