Publications

2026

1. ICPR2026

   

   Two-stage Vision Transformers and Hard Masking offer Robust Object Representations

   Ananthu Aniraj, Cassio F. Dantas, Dino Ienco, and Diego Marcos

   In International Conference on Pattern Recognition (ICPR), 2026

   Abs arXiv Bib PDF Code

   Context can strongly affect object representations, sometimes leading to undesired biases, particularly when objects appear in out-of-distribution backgrounds at inference. At the same time, many object-centric tasks require to leverage the context for identifying the relevant image regions. We posit that this conundrum, in which context is simultaneously needed and a potential nuisance, can be addressed by an attention-based approach that uses learned binary attention masks to ensure that only attended image regions influence the prediction. To test this hypothesis, we evaluate a two-stage framework: stage 1 processes the full image to discover object parts and identify task-relevant regions, for which context cues are likely to be needed, while stage 2 leverages input attention masking to restrict its receptive field to these regions, enabling a focused analysis while filtering out potentially spurious information. Both stages are trained jointly, allowing stage 2 to refine stage 1. The explicit nature of the semantic masks also makes the model’s reasoning auditable, enabling powerful test-time interventions to further enhance robustness. Extensive experiments across diverse benchmarks demonstrate that this approach significantly improves robustness against spurious correlations and out-of-distribution backgrounds.

   @inproceedings{aniraj2026ifam,
     title = {Two-stage Vision Transformers and Hard Masking offer Robust Object Representations},
     author = {Aniraj, Ananthu and Dantas, Cassio F. and Ienco, Dino and Marcos, Diego},
     booktitle = {International Conference on Pattern Recognition (ICPR)},
     year = {2026},
   }

2025

1. European-Patent

   

   Animal husbandry system

   Yan Li, Paulus Jacobus Maria Van Adrichem, Ananthu ANIRAJ, Cagla SOZEN, and 2 more authors

   Feb 2025

   Abs Bib HTML PDF

   The present invention relates to an animal husbandry system with monitoring and analyzing means configured for repeatedly generating images and gathering animal data therefrom. A plurality of cameras is suitable for monitoring the area from above. Image processing means are suitable for matching animals in the field of view of multiple cameras by detecting animal silhouettes in images of different cameras; defining an animal bounding box around each animal silhouette, applying to each animal bounding box either a first projection onto an image stitching plane if the animal is standing, or a second projection if the animal is lying down, stitching the projected bounding boxes in the stitching plane and identifying identical animals in the stitched image. The reliability of the matching process is increased by taking into account whether the animal in question is standing or lying down. The quality of the data gathered in the system is improved.

   @patent{li_animal_2023,
     title = {Animal husbandry system},
     url = {https://patents.google.com/patent/EP4510823A1/en},
     nationality = {EP},
     language = {en},
     assignee = {Lely Patent N.V.},
     number = {EP4510823A1},
     urldate = {2025-02-26},
     author = {Li, Yan and Adrichem, Paulus Jacobus Maria Van and ANIRAJ, Ananthu and SOZEN, Cagla and SLUIS, Joram Robin VAN DER and VLIET, Sjoerd Timo VAN},
     month = feb,
     year = {2025},
     keywords = {animal, animals, image, processing means, standing},
   }

2024

1. ECCV2024

   

   PDiscoFormer: Relaxing Part Discovery Constraints with Vision Transformers

   Ananthu Aniraj, Cassio F. Dantas, Dino Ienco, and Diego Marcos

   In European Conference on Computer Vision, Feb 2024

   Abs arXiv Bib HTML PDF Code

   Computer vision methods that explicitly detect object parts and reason on them are a step towards inherently interpretable models. Existing approaches that perform part discovery driven by a fine-grained classification task make very restrictive assumptions on the geometric properties of the discovered parts; they should be small and compact. Although this prior is useful in some cases, in this paper we show that pre-trained transformer-based vision models, such as self-supervised DINOv2 ViT, enable the relaxation of these constraints. In particular, we find that a total variation (TV) prior, which allows for multiple connected components of any size, substantially outperforms previous work. We test our approach on three fine-grained classification benchmarks: CUB, PartImageNet and Oxford Flowers, and compare our results to previously published methods as well as a re-implementation of the state-of-the-art method PDiscoNet with a transformer-based backbone. We consistently obtain substantial improvements across the board, both on part discovery metrics and the downstream classification task, showing that the strong inductive biases in self-supervised ViT models require to rethink the geometric priors that can be used for unsupervised part discovery.

   @inproceedings{aniraj2024pdiscoformer,
     title = {{PDiscoFormer: Relaxing Part Discovery Constraints with Vision Transformers}},
     author = {Aniraj, Ananthu and Dantas, Cassio F. and Ienco, Dino and Marcos, Diego},
     booktitle = {European Conference on Computer Vision},
     year = {2024},
     organization = {Springer},
     pages = {256-272},
     doi = {10.1007/978-3-031-73013-9\_15},
   }

2. European-Patent

   

   System for monitoring a calving mammal

   Adrianus Cornelis Maria MEEUWESEN, Yan Li, and Ananthu ANIRAJ

   Oct 2024

   Abs Bib HTML PDF

   The present invention relates to a calving monitoring system for monitoring an animal at the end of the expected gestation period, comprising a camera device for repeatedly taking images of the animal, said images being composed of pixels, a control unit for the calving monitoring system, connected to the camera device, which is configured to generate calving information from the images taken, and an alert device for sending an alert message according to the calving information generated, wherein the control unit is configured so as, in each image taken, to determine a parameter value relating to first pixels of said taken images as a time-dependent parametric function, and to detect contractions when said parameter value meets a predetermined contraction criterion, wherein the control unit generates calving information which comprises an indicator of the contractions detected.

   @patent{meeuwesen_system_2024,
     title = {System for monitoring a calving mammal},
     url = {https://patents.google.com/patent/EP4291133B1/en},
     nationality = {EP},
     assignee = {Lely Patent NV},
     number = {EP4291133B1},
     urldate = {2024-12-10},
     author = {MEEUWESEN, Adrianus Cornelis Maria and Li, Yan and ANIRAJ, Ananthu},
     month = oct,
     year = {2024},
     keywords = {animal, calving, contractions, control unit, torso},
   }

2023

1. ICCVW2023

   

   Masking Strategies for Background Bias Removal in Computer Vision Models

   Ananthu Aniraj, Cassio F. Dantas, Dino Ienco, and Diego Marcos

   In IEEE/CVF International Conference on Computer Vision Workshops (ICCVW), Oct 2023

   Abs arXiv Bib HTML PDF Code

   Models for fine-grained image classification tasks, where the difference between some classes can be extremely subtle and the number of samples per class tends to be low, are particularly prone to picking up background-related biases and demand robust methods to handle potential examples with out-of-distribution (OOD) backgrounds. To gain deeper insights into this critical problem, our research investigates the impact of background-induced bias on fine-grained image classification, evaluating standard backbone models such as Convolutional Neural Network (CNN) and Vision Transformers (ViT). We explore two masking strategies to mitigate background-induced bias: Early masking, which removes background information at the (input) image level, and late masking, which selectively masks high-level spatial features corresponding to the background. Extensive experiments assess the behavior of CNN and ViT models under different masking strategies, with a focus on their generalization to OOD backgrounds. The obtained findings demonstrate that both proposed strategies enhance OOD performance compared to the baseline models, with early masking consistently exhibiting the best OOD performance. Notably, a ViT variant employing GAP-Pooled Patch token-based classification combined with early masking achieves the highest OOD robustness.

   @inproceedings{10351024,
     author = {Aniraj, Ananthu and Dantas, Cassio F. and Ienco, Dino and Marcos, Diego},
     booktitle = {IEEE/CVF International Conference on Computer Vision Workshops (ICCVW)},
     title = {Masking Strategies for Background Bias Removal in Computer Vision Models},
     year = {2023},
     volume = {},
     issn = {},
     pages = {4399-4407},
     keywords = {computer vision;computational modeling;transformers;data models;robustness;convolutional neural networks;task analysis},
     doi = {10.1109/ICCVW60793.2023.00474},
     url = {https://doi.ieeecomputersociety.org/10.1109/ICCVW60793.2023.00474},
     publisher = {IEEE Computer Society},
     address = {Los Alamitos, CA, USA},
     month = oct,
   }