Andrew D. Bagdanov — accepted AI-conference papers

2026

IsoCLIP: Decomposing CLIP Projectors for Efficient Intra-modal Alignment

Simone Magistri, Dipam Goswami, Marco Mistretta, Bartłomiej Twardowski, Joost van de Weijer, Andrew D. Bagdanov

CVPR 2026

Vision-Language Models like CLIP are extensively used for inter-modal tasks which involve both visual and text modalities. However, when the individual modality encoders are applied to inherently intra-modal tasks like image-to-image retrieval, their performance suffers from the intra-modal misalign

Cited by 0 Source ↗code ↗Scholar ↗

2026

SpectralGCD: Spectral Concept Selection and Cross-modal Representation Learning for Generalized Category Discovery

Lorenzo Caselli, Marco Mistretta, Simone Magistri, Andrew D. Bagdanov

ICLR 2026poster

Generalized Category Discovery (GCD) aims to identify novel categories in unlabeled data while leveraging a small labeled subset of known classes. Training a parametric classifier solely on image features often leads to overfitting to old classes, and recent multimodal approaches improve performance…

Cited by 0 Source ↗code ↗Scholar ↗

2025

Accurate and Efficient Low-Rank Model Merging in Core Space

Aniello Panariello, Daniel Marczak, Simone Magistri, Angelo Porrello, Bartłomiej Twardowski, Andrew D. Bagdanov +2

NeurIPS 2025poster

In this paper, we address the challenges associated with merging low-rank adaptations of large neural networks. With the rise of parameter-efficient adaptation techniques, such as Low-Rank Adaptation (LoRA), model fine-tuning has become more accessible. While fine-tuning models with LoRA is highly e…

Cited by 0 Source ↗code ↗Scholar ↗

2025

Covariances for Free: Exploiting Mean Distributions for Training-free Federated Learning

Dipam Goswami, Simone Magistri, Kai Wang, Bartłomiej Twardowski, Andrew D. Bagdanov, Joost van de Weijer

NeurIPS 2025poster

Using pre-trained models has been found to reduce the effect of data heterogeneity and speed up federated learning algorithms. Recent works have explored training-free methods using first- and second-order statistics to aggregate local client data distributions at the server and achieve high perform…

Cited by 0 Source ↗code ↗Scholar ↗

2025

Cross the Gap: Exposing the Intra-modal Misalignment in CLIP via Modality Inversion

Marco Mistretta, Alberto Baldrati, Lorenzo Agnolucci, Marco Bertini, Andrew D. Bagdanov

ICLR 2025poster

Pre-trained multi-modal Vision-Language Models like CLIP are widely used off-the-shelf for a variety of applications. In this paper, we show that the common practice of individually exploiting the text or image encoders of these powerful multi-modal models is highly suboptimal for intra-modal tasks…

Cited by 1 Source ↗PDF ↗code ↗Scholar ↗

2025

No Task Left Behind: Isotropic Model Merging with Common and Task-Specific Subspaces

Daniel Marczak, Simone Magistri, Sebastian Cygert, Bartłomiej Twardowski, Andrew D. Bagdanov, Joost van de Weijer

ICML 2025poster

Model merging integrates the weights of multiple task-specific models into a single multi-task model. Despite recent interest in the problem, a significant performance gap between the combined and single-task models remains. In this paper, we investigate the key characteristics of task matrices --…

Cited by 1 Source ↗PDF ↗code ↗Scholar ↗

2024

Elastic Feature Consolidation For Cold Start Exemplar-Free Incremental Learning

Simone Magistri, Tomaso Trinci, Albin Soutif, Joost van de Weijer, Andrew D. Bagdanov

ICLR 2024poster

Exemplar-Free Class Incremental Learning (EFCIL) aims to learn from a sequence of tasks without having access to previous task data. In this paper, we consider the challenging Cold Start scenario in which insufficient data is available in the first task to learn a high-quality backbone. This is espe…

Cited by 23 Source ↗PDF ↗code ↗Scholar ↗

2024

Improving Zero-shot Generalization of Learned Prompts via Unsupervised Knowledge Distillation

Marco Mistretta*, Alberto Baldrati, Marco Bertini, Andrew D. Bagdanov

ECCV 2024poster

"Vision-Language Models (VLMs) demonstrate remarkable zero-shot generalization to unseen tasks, but fall short of the performance of supervised methods in generalizing to downstream tasks with limited data. Prompt learning is emerging as a parameter-efficient method for adapting VLMs, but state-of-t…

Cited by 7 Source ↗PDF ↗code ↗Scholar ↗

2024

Task-Adaptive Saliency Guidance for Exemplar-free Class Incremental Learning

Xialei Liu, Jiang-Tian Zhai, Andrew D. Bagdanov, Ke Li, Ming-Ming Cheng

CVPR 2024poster

Exemplar-free Class Incremental Learning (EFCIL) aims to sequentially learn tasks with access only to data from the current one. EFCIL is of interest because it mitigates concerns about privacy and long-term storage of data while at the same time alleviating the problem of catastrophic forgetting in…

Cited by 5 Source ↗PDF ↗code ↗Scholar ↗

2023

Masked Autoencoders are Efficient Class Incremental Learners

Jiang-Tian Zhai, Xialei Liu, Andrew D. Bagdanov, Ke Li, Ming-Ming Cheng

ICCV 2023poster

Class Incremental Learning (CIL) aims to sequentially learn new classes while avoiding catastrophic forgetting of previous knowledge. We propose to use Masked Autoencoders (MAEs) as efficient learners for CIL. MAEs were originally designed to learn useful representations through reconstr…

Cited by 17 PDF ↗code ↗Scholar ↗

2023

Planckian Jitter: countering the color-crippling effects of color jitter on self-supervised training

Simone Zini, Alex Gomez-Villa, Marco Buzzelli, Bartłomiej Twardowski, Andrew D. Bagdanov, Joost van de weijer

ICLR 2023poster

Several recent works on self-supervised learning are trained by mapping different augmentations of the same image to the same feature representation. The data augmentations used are of crucial importance to the quality of learned feature representations. In this paper, we analyze how the color jitte…

Cited by 15 Source ↗PDF ↗code ↗Scholar ↗

2022

Long-Tailed Class Incremental Learning

Xialei Liu, Yu-Song Hu, Xu-Sheng Cao, Andrew D. Bagdanov, Ke Li, Ming-Ming Cheng

ECCV 2022poster

"In class incremental learning (CIL) a model must learn new classes in a sequential manner without forgetting old ones. However, conventional CIL methods consider a balanced distribution for each new task, which ignores the prevalence of long-tailed distributions in the real world. In this work we p…

Cited by 53 Source ↗PDF ↗code ↗Scholar ↗

2020

RATT: Recurrent Attention to Transient Tasks for Continual Image Captioning

Riccardo Del Chiaro, Bartłomiej Twardowski, Andrew D. Bagdanov, Joost van de Weijer

NeurIPS 2020poster

Research on continual learning has led to a variety of approaches to mitigating catastrophic forgetting in feed-forward classification networks. Until now surprisingly little attention has been focused on continual learning of recurrent models applied to problems like image captioning. In this paper…

Cited by 59 Source ↗PDF ↗code ↗Scholar ↗

2020

Task-conditioned Domain Adaptation for Pedestrian Detection in Thermal Imagery

My Kieu, Andrew D. Bagdanov, Marco Bertini, Alberto del Bimbo

ECCV 2020poster

Pedestrian detection is a core problem in computer vision that sees broad application in video surveillance and, more recently, in advanced driving assistance systems. Despite its broad application and interest, it remains a challenging problem in part due to the vast range of conditions under which…

Cited by 72 Source ↗PDF ↗code ↗Scholar ↗

2018

Leveraging Unlabeled Data for Crowd Counting by Learning to Rank

Xialei Liu, Joost van de Weijer, Andrew D. Bagdanov

CVPR 2018poster

We propose a novel crowd counting approach that leverages abundantly available unlabeled crowd imagery in a learning-to-rank framework. To induce a ranking of cropped images , we use the observation that any sub-image of a crowded scene image is guaranteed to contain the same number or fewer pe…

Cited by 342 Source ↗PDF ↗code ↗Scholar ↗

2017

Domain-Adaptive Deep Network Compression

Marc Masana, Joost van de Weijer, Luis Herranz, Andrew D. Bagdanov, Jose M. Alvarez

ICCV 2017poster

Deep Neural Networks trained on large datasets can be easily transferred to new domains with far fewer labeled examples by a process called fine-tuning. This has the advantage that representations learned in the large source domain can be exploited on smaller target domains. However, networks design…

Cited by 80 PDF ↗code ↗Scholar ↗

2017

RankIQA: Learning From Rankings for No-Reference Image Quality Assessment

Xialei Liu, Joost van de Weijer, Andrew D. Bagdanov

ICCV 2017poster

We propose a no-reference image quality assessment (NR-IQA) approach that learns from rankings (RankIQA). To address the problem of limited IQA dataset size, we train a Siamese Network to rank images in terms of image quality by using synthetically generated distortions for which relative image qual…

Cited by 591 PDF ↗code ↗Scholar ↗