Francisco Caetano
PhD Candidate · Eindhoven University of Technology
My research focuses on practical, deployable generative solutions for image generation and editing, domain adaptation, and out-of-distribution detection to tackle real-world problems in medical imaging. I am now extending this work into unifying generative and discriminative frameworks.
Publications
Symmetrical Flow Matching: Unified Image Generation, Segmentation, and Classification with Score-Based Generative Models
Francisco Caetano, C. Viviers, P.H.N. de With, F. van der Sommen
AAAI 2026
[
Project Page | Paper | Code
]
DisCoPatch: Taming Adversarially-driven Batch Statistics for Improved Out-of-Distribution Detection
Francisco Caetano, C. Viviers, L. Mondragon, P.H.N. de With, F. van der Sommen
ICCV 2025
[
Project Page | Paper | Code
]
MedShift: Implicit Conditional Transport for X-Ray Domain Adaptation
Francisco Caetano, C. Viviers, P.H.N. de With, F. van der Sommen
ICCV 2025 Workshop
[
Project Page | Paper | Code
]
MedSymmFlow: Bridging Generative Modeling and Classification in Medical Imaging through Symmetrical Flow Matching
Francisco Caetano, L. Abdi, C. Viviers, A. Valiuddin, F. van der Sommen
MICCAI 2025 Workshop
[
Project Page | Paper | Code
]
AdverX-Ray: Ensuring X-Ray Integrity Through Frequency-Sensitive Adversarial VAEs
Francisco Caetano, C. Viviers, L. Filatova, P.H.N. de With, F. van der Sommen
SPIE Medical Imaging 2025
[
Project Page | Paper | Code
]
Runner-up · Robert F. Wagner All-Conference Best Student Paper Award
Can Your Generative Model Detect Out-of-Distribution Covariate Shift?
C. Viviers, A. Valiuddin, Francisco Caetano, L. Abdi, L. Filatova, P.H.N. de With, F. van der Sommen
ECCV 2024 Workshop
[
Paper | Code
]
Robust Early Detection of Barrett’s Neoplasia: Addressing Low-Prevalence Challenges with Generative Modeling
T.J.M. Jaspers, Francisco Caetano, C.H.B. Claessens, C.H.J. Kusters, H. Middeljans, M.R. Jong, R.A.H. van Eijck van Heslinga, F. Slooter, A.J. de Groof, J.J. Bergman, P.H.N. De With, F. van der Sommen
MICCAI 2025 Workshop
[
Paper
]
Other publications & reviews
Zero-Shot Image Anomaly Detection Using Generative Foundation Models
L. Abdi, A. Valiuddin, Francisco Caetano, C. Viviers, F. van der Sommen
ICCV 2025 Workshop
[
Paper
]
Out-of-Distribution Detection in Medical Imaging via Diffusion Trajectories
L. Abdi, Francisco Caetano, A. Valiuddin, C. Viviers, H. Joudeh, F. van der Sommen
MICCAI 2025 Workshop
[
Paper
]
Diffusion-based Lung Nodule Synthesis for Advanced Evaluation of Deep Learning Models
C.H.B. Claessens, Francisco Caetano, K. van der Wulp, L.J.S. Ewals, F. van der Sommen
SPIE Medical Imaging 2025
[
Paper | Code
]
Visual Data Processing for Anomaly Detection
Francisco Caetano
Master Thesis
[
Paper
]
Enhancing Weakly-Supervised Video Anomaly Detection with Temporal Constraints
Francisco Caetano, P. Carvalho, C. Mastralexi, J. Cardoso
IEEE Access
[
Paper
]
Unveiling the Performance of Video Anomaly Detection Models — a Benchmark-based Review
Francisco Caetano, P. Carvalho, J. Cardoso
Intelligent Systems with Applications
[
Paper
]
Deep Anomaly Detection for In-Vehicle Monitoring — An Application-Oriented Review
Francisco Caetano, P. Carvalho, J. Cardoso
Applied Sciences (MDPI)
[
Paper
]
Projects
GenerativeZoo — Unified repository of generative models for computer vision, providing reproducible baselines and tooling for the research community. [ Project Page | GitHub ]
RARE25 — The RARE25 Challenge focuses on building a classification system that can accurately detect early-stage cancer in patients with Barrett’s Esophagus. [ Project Page ]
Experience
PhD Candidate
Eindhoven University of Technology
- Generative artificial intelligence for tailored synthetic image generation
- Supervised by dr.ir. Fons van der Sommen
Computer Vision Researcher
Fraunhofer Portugal AICOS
- Anomaly detection for in-line visual inspection
MSc Electrical Engineering
Faculty of Engineering, University of Porto
- Graduated with GPA of 18/20
- Published two journal articles during the Master thesis
Research Assistant
INESC TEC - CTM
- Handled occlusion-aware recognition of human activity
- Built real-time trajectory prediction for surrounding vehicles
BSc Electrical Engineering
Faculty of Engineering, University of Porto
- Graduated with GPA of 18/20