2606.09670v1 Jun 08, 2026 cs.CV

Visual Prompting Meets Feature Reconstruction-Based Anomaly Detection with Dual-Teacher Supervision

Niccolo Avogaro
Niccolo Avogaro
Citations: 9
h-index: 2
Thomas Frick
Thomas Frick
Citations: 16
h-index: 2
Mattia Rigotti
Mattia Rigotti
IBM Research AI
Citations: 5,157
h-index: 23
Daniel Caraballo
Daniel Caraballo
IBM
Citations: 4
h-index: 2
Y. Çinar
Y. Çinar
Citations: 5
h-index: 2
F. Scheidegger
F. Scheidegger
Citations: 10
h-index: 2
Andrea Bartezzaghi
Andrea Bartezzaghi
Citations: 0
h-index: 0
Cezary Skura
Cezary Skura
Citations: 2
h-index: 1
Brown Ebouky
Brown Ebouky
Citations: 11
h-index: 2
Mateo Diaz-Bone
Mateo Diaz-Bone
Citations: 0
h-index: 0
Roy Assaf
Roy Assaf
Citations: 9
h-index: 2
Filip M. Janicki
Filip M. Janicki
Citations: 10
h-index: 2
Piotr Kluska
Piotr Kluska
Citations: 7
h-index: 2
C. Malossi
C. Malossi
Citations: 135
h-index: 4

Recent Anomaly Detection methods achieve perfect detection and segmentation scores on well-established datasets, such as MVTec. However, many of these methods face challenges when foundational assumptions - such as consistent object scale, viewpoint, background, illumination, and centered placement - are violated. Those variations that occur render anomaly detection methods unusable in many real-world scenarios. To address these limitations, we introduce three key contributions: (1) a visual prompting pipeline that isolates objects using foreground-background masking; (2) a mechanism for unfreezing the teacher in student-teacher models to improve domain adaptability; and (3) a data augmentation strategy leveraging diffusion-generated synthetic images to enhance anomaly detection performance. We achieve a 3.5 percentage point improvement over the previous state-of-the-art on the challenging AeBAD dataset by using the Masked Multiscale Reconstruction (MMR) model as our backbone.

0 Citations
0 Influential
11.5 Altmetric
57.5 Score
Original PDF

No Analysis Report Yet

This paper hasn't been analyzed by Gemini yet.

Log in to request an AI analysis.

댓글

댓글을 작성하려면 로그인하세요.

아직 댓글이 없습니다. 첫 번째 댓글을 남겨보세요!