CA-Stream: Attention-based pooling for interpretable image recognition - Archive ouverte HAL Access content directly
Conference Papers Year : 2024

CA-Stream: Attention-based pooling for interpretable image recognition

Felipe Torres
  • Function : Author
  • PersonId : 1376426
Hanwei Zhang
  • Function : Author
  • PersonId : 1376427
Ronan Sicre
  • Function : Author
  • PersonId : 1067988
Stéphane Ayache

Abstract

Explanations obtained from transformer-based architectures in the form of raw attention, can be seen as a class-agnostic saliency map. Additionally, attention-based pooling serves as a form of masking the in feature space. Motivated by this observation, we design an attention-based pooling mechanism intended to replace Global Average Pooling (GAP) at inference. This mechanism, called Cross-Attention Stream (CA-Stream), comprises a stream of cross attention blocks interacting with features at different network depths. CA-Stream enhances interpretability in models, while preserving recognition performance.
Fichier principal
Vignette du fichier
WACV_ICCVround2-4.pdf (2.29 Mo) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-04551613 , version 1 (18-04-2024)

Identifiers

  • HAL Id : hal-04551613 , version 1

Cite

Felipe Torres, Hanwei Zhang, Ronan Sicre, Stéphane Ayache, Yannis Avrithis. CA-Stream: Attention-based pooling for interpretable image recognition. XAI4CV workshop (CVPR), Jun 2024, Seatle, WA, United States. ⟨hal-04551613⟩
21 View
32 Download

Share

Gmail Mastodon Facebook X LinkedIn More