2 code implementations • 12 Apr 2024 • Sina Hajimiri, Ismail Ben Ayed, Jose Dolz
However, existing approaches often rely on impractical supervised pre-training or access to additional pre-trained networks.
1 code implementation • 2 Apr 2024 • Yunshi Huang, Fereshteh Shakeri, Jose Dolz, Malik Boudiaf, Houda Bahig, Ismail Ben Ayed
In this work, we propose and examine from convex-optimization perspectives a generalization of the standard LP baseline, in which the linear classifier weights are learnable functions of the text embedding, with class-wise multipliers blending image and text knowledge.
1 code implementation • 22 Mar 2024 • Shambhavi Mishra, Balamurali Murugesan, Ismail Ben Ayed, Marco Pedersoli, Jose Dolz
State-of-the-art semi-supervised learning (SSL) approaches rely on highly confident predictions to serve as pseudo-labels that guide the training on unlabeled samples.
1 code implementation • 19 Mar 2024 • Balamurali Murugesan, Julio Silva-Rodriguez, Ismail Ben Ayed, Jose Dolz
In particular, we present a formulation that integrates class and region-wise constraints into the learning objective, with multiple penalty weights to account for class and region differences.
no code implementations • 27 Jan 2024 • Julio Silva-Rodriguez, Jihed Chelbi, Waziha Kabir, Hadi Chakor, Jose Dolz, Ismail Ben Ayed, Riadh Kobbi
In this work, we explore the potential of using FLAIR features as starting point for fundus image classification, and we compare its performance with regard to Imagenet initialization on two popular transfer learning methods: Linear Probing (LP) and Fine-Tuning (FP).
no code implementations • 25 Jan 2024 • Balamurali Murugesan, Sukesh Adiga Vasudeva, Bingyuan Liu, Hervé Lombaert, Ismail Ben Ayed, Jose Dolz
Ensuring reliable confidence scores from deep neural networks is of paramount significance in critical decision-making systems, particularly in real-world domains such as healthcare.
1 code implementation • 20 Dec 2023 • Julio Silva-Rodríguez, Sina Hajimiri, Ismail Ben Ayed, Jose Dolz
Efficient transfer learning (ETL) is receiving increasing attention to adapt large pre-trained language-vision models on downstream tasks with a few labeled samples.
1 code implementation • 24 Oct 2023 • Sukesh Adiga V, Jose Dolz, Herve Lombaert
This work proposes a novel method to estimate segmentation uncertainty by leveraging global information from the segmentation masks.
1 code implementation • 15 Aug 2023 • Julio Silva-Rodriguez, Hadi Chakor, Riadh Kobbi, Jose Dolz, Ismail Ben Ayed
Foundation vision-language models are currently transforming computer vision, and are on the rise in medical imaging fueled by their very promising generalization capabilities.
no code implementations • 11 Jul 2023 • Julien Nicolas, Florent Chiaroni, Imtiaz Ziko, Ola Ahmad, Christian Desrosiers, Jose Dolz
Despite the recent progress in incremental learning, addressing catastrophic forgetting under distributional drift is still an open and important problem.
2 code implementations • 30 Jun 2023 • Balamurali Murugesan, Rukhshanda Hussain, Rajarshi Bhattacharya, Ismail Ben Ayed, Jose Dolz
First, modifying only the class token of the text prompt results in a greater impact on the Class Activation Map (CAM), compared to arguably more complex strategies that optimize the context.
Few-Shot Learning Weakly supervised Semantic Segmentation +1
no code implementations • 23 May 2023 • Bach Kim, Jose Dolz, Pierre-Marc Jodoin, Christian Desrosiers
Our system has two components: 1) a segmentation network on the server side which processes the image mixture, and 2) a segmentation unmixing network which recovers the correct segmentation map from the segmentation mixture.
1 code implementation • 29 Mar 2023 • Julio Silva-Rodríguez, Jose Dolz, Ismail Ben Ayed
With the recent raise of foundation models in computer vision and NLP, the pretrain-and-adapt strategy, where a large-scale model is fine-tuned on downstream tasks, is gaining popularity.
1 code implementation • 11 Mar 2023 • Balamurali Murugesan, Sukesh Adiga V, Bingyuan Liu, Hervé Lombaert, Ismail Ben Ayed, Jose Dolz
Ensuring reliable confidence scores from deep networks is of pivotal importance in critical decision-making systems, notably in the medical domain.
1 code implementation • 27 Jan 2023 • Farzad Beizaee, Christian Desrosiers, Gregory A. Lodygensky, Jose Dolz
In this paper, we propose an unsupervised framework based on normalizing flows that harmonizes MR images to mimic the distribution of the source domain.
1 code implementation • ICCV 2023 • Florent Chiaroni, Jose Dolz, Ziko Imtiaz Masud, Amar Mitiche, Ismail Ben Ayed
We introduce a Parametric Information Maximization (PIM) model for the Generalized Category Discovery (GCD) problem.
1 code implementation • CVPR 2023 • Bingyuan Liu, Jérôme Rony, Adrian Galdran, Jose Dolz, Ismail Ben Ayed
Comprehensive evaluation and multiple comparisons on a variety of benchmarks, including standard and long-tailed image classification, semantic segmentation, and text classification, demonstrate the superiority of the proposed method.
1 code implementation • CVPR 2023 • Sina Hajimiri, Malik Boudiaf, Ismail Ben Ayed, Jose Dolz
In addition, the terms derived from our MI-based formulation are coupled with a knowledge distillation term to retain the knowledge on base classes.
1 code implementation • 9 Sep 2022 • Balamurali Murugesan, Bingyuan Liu, Adrian Galdran, Ismail Ben Ayed, Jose Dolz
Following our observations, we propose a simple and flexible generalization based on inequality constraints, which imposes a controllable margin on logit distances.
no code implementations • 18 Jun 2022 • Sukesh Adiga V, Jose Dolz, Herve Lombaert
This integrated attention mechanism provides a visual insight of discriminative image features that contribute to the clustering of image sets and a visual explanation of the embedding features.
1 code implementation • 12 May 2022 • Soufiane Belharbi, Jérôme Rony, Jose Dolz, Ismail Ben Ayed, Luke McCaffrey, Eric Granger
Our method is composed of two networks: a localizer that yields segmentation mask, followed by a classifier.
1 code implementation • 10 Mar 2022 • Sukesh Adiga V, Jose Dolz, Herve Lombaert
The learnt labeling representation is used to map the prediction of the segmentation into a set of plausible masks.
no code implementations • 7 Mar 2022 • Martin Van Waerebeke, Gregory Lodygensky, Jose Dolz
Semi-supervised learning has emerged as an appealing strategy to train deep models with limited supervision.
1 code implementation • 3 Mar 2022 • Julio Silva-Rodríguez, Valery Naranjo, Jose Dolz
In particular, the equality constraint on the attention maps in prior work is replaced by an inequality constraint, which allows more flexibility.
no code implementations • 22 Dec 2021 • Agostina Larrazabal, Cesar Martinez, Jose Dolz, Enzo Ferrante
Modern deep neural networks achieved remarkable progress in medical image segmentation tasks.
1 code implementation • CVPR 2022 • Bingyuan Liu, Ismail Ben Ayed, Adrian Galdran, Jose Dolz
Following our observations, we propose a simple and flexible generalization based on inequality constraints, which imposes a controllable margin on logit distances.
2 code implementations • 21 Sep 2021 • Bingyuan Liu, Christian Desrosiers, Ismail Ben Ayed, Jose Dolz
Combined with a standard cross-entropy loss over the labeled pixels, our novel formulation integrates two important terms: (i) a Shannon entropy loss defined over the less-supervised images, which encourages confident student predictions in the bottom branch; and (ii) a KL divergence term, which transfers the knowledge (i. e., predictions) of the strongly supervised branch to the less-supervised branch and guides the entropy (student-confidence) term to avoid trivial solutions.
1 code implementation • 1 Sep 2021 • Julio Silva-Rodríguez, Valery Naranjo, Jose Dolz
In particular, the equality constraint on the attention maps in prior work is replaced by an inequality constraint, which allows more flexibility.
1 code implementation • 6 Aug 2021 • Mathilde Bateson, Hoel Kervadec, Jose Dolz, Hervé Lombaert, Ismail Ben Ayed
Our method yields comparable results to several state of the art adaptation techniques, despite having access to much less information, as the source images are entirely absent in our adaptation phase.
3 code implementations • 23 Jun 2021 • Malik Boudiaf, Ziko Imtiaz Masud, Jérôme Rony, Jose Dolz, Ismail Ben Ayed, Pablo Piantanida
We motivate our transductive loss by deriving a formal relation between the classification accuracy and mutual-information maximization.
1 code implementation • 16 Jun 2021 • Imtiaz Masud Ziko, Malik Boudiaf, Jose Dolz, Eric Granger, Ismail Ben Ayed
Surprisingly, we found that even standard clustering procedures (e. g., K-means), which correspond to particular, non-regularized cases of our general model, already achieve competitive performances in comparison to the state-of-the-art in few-shot learning.
1 code implementation • 22 May 2021 • Agostina J. Larrazabal, César Martínez, Jose Dolz, Enzo Ferrante
Despite the astonishing performance of deep-learning based approaches for visual tasks such as semantic segmentation, they are known to produce miscalibrated predictions, which could be harmful for critical decision-making processes.
1 code implementation • 21 May 2021 • Julio Silva-Rodríguez, Adrián Colomer, Jose Dolz, Valery Naranjo
Particularly, the proposed model brings an average improvement on the Cohen's quadratic kappa (k) score of nearly 18% compared to full-supervision for the patch-level Gleason grading task.
1 code implementation • 3 May 2021 • Hoel Kervadec, Houda Bahig, Laurent Letourneau-Guillon, Jose Dolz, Ismail Ben Ayed
We also found that shape descriptors can be a valid way to encode anatomical priors about the task, enabling to leverage expert knowledge without additional annotations.
1 code implementation • 18 Apr 2021 • Bingyuan Liu, Jose Dolz, Adrian Galdran, Riadh Kobbi, Ismail Ben Ayed
Most segmentation losses are arguably variants of the Cross-Entropy (CE) or Dice losses.
1 code implementation • 13 Apr 2021 • Le Thanh Nguyen-Meidine, Madhu Kiran, Marco Pedersoli, Jose Dolz, Louis-Antoine Blais-Morin, Eric Granger
Recent advances in unsupervised domain adaptation have significantly improved the recognition accuracy of CNNs by alleviating the domain shift between (labeled) source and (unlabeled) target data distributions.
1 code implementation • 6 Apr 2021 • Gaurav Patel, Jose Dolz
In addition, we add a KL-divergence on the class prediction distributions to facilitate the information exchange between modalities, which, combined with the equivariant regularizers further improves the performance of our model.
no code implementations • 18 Jan 2021 • Le Thanh Nguyen-Meidine, Atif Belal, Madhu Kiran, Jose Dolz, Louis-Antoine Blais-Morin, Eric Granger
Our proposed approach is compared against state-of-the-art methods for compression and STDA of CNNs on the Office31 and ImageClef-DA image classification datasets.
no code implementations • 16 Jan 2021 • Mark G. Bandyk, Dheeraj R Gopireddy, Chandana Lall, K. C. Balaji, Jose Dolz
Nevertheless, despite the success of these models in other medical problems, progress in multi region bladder segmentation is still at a nascent stage, with just a handful of works tackling a multi region scenario.
1 code implementation • 15 Dec 2020 • Jose Dolz, Christian Desrosiers, Ismail Ben Ayed
In conjunction with a standard cross-entropy over the labeled pixels, our novel formulation integrates two important terms: (i) a Shannon entropy loss defined over the less-supervised images, which encourages confident student predictions at the bottom branch; and (ii) a Kullback-Leibler (KL) divergence, which transfers the knowledge from the predictions generated by the strongly supervised branch to the less-supervised branch, and guides the entropy (student-confidence) term to avoid trivial solutions.
2 code implementations • CVPR 2021 • Malik Boudiaf, Hoel Kervadec, Ziko Imtiaz Masud, Pablo Piantanida, Ismail Ben Ayed, Jose Dolz
We show that the way inference is performed in few-shot segmentation tasks has a substantial effect on performances -- an aspect often overlooked in the literature in favor of the meta-learning paradigm.
Ranked #3 on Few-Shot Semantic Segmentation on COCO-20i (10-shot)
1 code implementation • NeurIPS 2020 • Malik Boudiaf, Imtiaz Ziko, Jérôme Rony, Jose Dolz, Pablo Piantanida, Ismail Ben Ayed
We introduce Transductive Infomation Maximization (TIM) for few-shot learning.
no code implementations • 25 Nov 2020 • Bach Ngoc Kim, Jose Dolz, Christian Desrosiers, Pierre-Marc Jodoin
Results show that the segmentation accuracy of our method is similar to a system trained on non-encoded images, while considerably reducing the ability to recover subject identity.
2 code implementations • 14 Nov 2020 • Soufiane Belharbi, Jérôme Rony, Jose Dolz, Ismail Ben Ayed, Luke McCaffrey, Eric Granger
We propose novel regularization terms, which enable the model to seek both non-discriminative and discriminative regions, while discouraging unbalanced segmentations.
1 code implementation • 14 Jul 2020 • Le Thanh Nguyen-Meidine, Atif Belal, Madhu Kiran, Jose Dolz, Louis-Antoine Blais-Morin, Eric Granger
Unsupervised domain adaptation (UDA) seeks to alleviate the problem of domain shift between the distribution of unlabeled data from the target domain w. r. t.
Ranked #3 on Multi-target Domain Adaptation on Office-31
2 code implementations • 28 Jun 2020 • Imtiaz Masud Ziko, Jose Dolz, Eric Granger, Ismail Ben Ayed
Our transductive inference does not re-train the base model, and can be viewed as a graph clustering of the query set, subject to supervision constraints from the support set.
2 code implementations • 16 May 2020 • Le Thanh Nguyen-Meidine, Eric Granger, Madhu Kiran, Jose Dolz, Louis-Antoine Blais-Morin
In both datasets, results indicate that our method can achieve the highest level of accuracy while requiring a comparable or lower time complexity.
5 code implementations • 7 May 2020 • Mathilde Bateson, Hoel Kervadec, Jose Dolz, Herve Lombaert, Ismail Ben Ayed
Our formulation is based on minimizing a label-free entropy loss defined over target-domain data, which we further guide with a domain invariant prior on the segmentation regions.
1 code implementation • MIDL 2019 • Hoel Kervadec, Jose Dolz, Shan-Shan Wang, Eric Granger, Ismail Ben Ayed
Particularly, we leverage a classical tightness prior to a deep learning setting via imposing a set of constraints on the network outputs.
no code implementations • 7 Apr 2020 • Sukesh Adiga V, Jose Dolz, Herve Lombaert
Segmentation using deep learning has shown promising directions in medical imaging as it aids in the analysis and diagnosis of diseases.
no code implementations • 18 Mar 2020 • Abdur R Feyjie, Reza Azad, Marco Pedersoli, Claude Kauffman, Ismail Ben Ayed, Jose Dolz
To handle this new learning paradigm, we propose to include surrogate tasks that can leverage very powerful supervisory signals --derived from the data itself-- for semantic feature learning.
1 code implementation • 9 Mar 2020 • Reza Azad, Abdur R Fayjie, Claude Kauffman, Ismail Ben Ayed, Marco Pedersoli, Jose Dolz
Despite the initial belief that Convolutional Neural Networks (CNNs) are driven by shapes to perform visual recognition tasks, recent evidence suggests that texture bias in CNNs provides higher performing models when learning on large labeled training datasets.
Ranked #2 on Few-Shot Semantic Segmentation on Pascal5i
no code implementations • 9 Sep 2019 • Bach Ngoc Kim, Jose Dolz, Pierre-Marc Jodoin, Christian Desrosiers
Our novel architecture is composed of three components: 1) an encoder network which removes identity-specific features from input medical images, 2) a discriminator network that attempts to identify the subject from the encoded images, 3) a medical image analysis network which analyzes the content of the encoded images (segmentation in our case).
1 code implementation • 8 Sep 2019 • Jérôme Rony, Soufiane Belharbi, Jose Dolz, Ismail Ben Ayed, Luke McCaffrey, Eric Granger
Four key challenges are identified for the application of deep WSOL methods in histology -- under/over activation of CAMs, sensitivity to thresholding, and model selection.
1 code implementation • 30 Aug 2019 • Arnab Kumar Mondal, Aniket Agarwal, Jose Dolz, Christian Desrosiers
In this work, we study the problem of training deep networks for semantic image segmentation using only a fraction of annotated images, which may significantly reduce human annotation efforts.
no code implementations • 15 Aug 2019 • Jizong Peng, Hoel Kervadec, Jose Dolz, Ismail Ben Ayed, Marco Pedersoli, Christian Desrosiers
An efficient strategy for weakly-supervised segmentation is to impose constraints or regularization priors on target regions.
1 code implementation • 8 Aug 2019 • Mathilde Bateson, Jose Dolz, Hoel Kervadec, Hervé Lombaert, Ismail Ben Ayed
We propose to adapt segmentation networks with a constrained formulation, which embeds domain-invariant prior knowledge about the segmentation regions.
1 code implementation • 25 Jul 2019 • Soufiane Belharbi, Jérôme Rony, Jose Dolz, Ismail Ben Ayed, Luke McCaffrey, Eric Granger
Pointwise localization allows more precise localization and accurate interpretability, compared to bounding box, in applications where objects are highly unstructured such as in medical domain.
1 code implementation • arXiv preprint 2019 • Ashish Sinha, Jose Dolz
In this paper we attempt to overcome these limitations with the proposed architecture, by capturing richer contextual dependencies based on the use of guided self-attention mechanisms.
Ranked #1 on Medical Image Segmentation on HSVM
1 code implementation • 10 Apr 2019 • Hoel Kervadec, Jose Dolz, Eric Granger, Ismail Ben Ayed
This study investigates a curriculum-style strategy for semi-supervised CNN segmentation, which devises a regression network to learn image-level information such as the size of a target region.
1 code implementation • 8 Apr 2019 • Hoel Kervadec, Jose Dolz, Jing Yuan, Christian Desrosiers, Eric Granger, Ismail Ben Ayed
While sub-optimality is not guaranteed for non-convex problems, this result shows that log-barrier extensions are a principled way to approximate Lagrangian optimization for constrained CNNs via implicit dual variables.
no code implementations • MIDL 2019 • Georg Pichler, Jose Dolz, Ismail Ben Ayed, Pablo Piantanida
We juxtapose our approach to state-of-the-art segmentation adaptation via adversarial training in the network-output space.
5 code implementations • 17 Dec 2018 • Hoel Kervadec, Jihene Bouchtiba, Christian Desrosiers, Eric Granger, Jose Dolz, Ismail Ben Ayed
We propose a boundary loss, which takes the form of a distance metric on the space of contours, not regions.
Brain Lesion Segmentation From Mri Ischemic Stroke Lesion Segmentation +4
1 code implementation • 19 Nov 2018 • Jose Dolz, Christian Desrosiers, Ismail Ben Ayed
Despite the technological advances in medical imaging, IVD localization and segmentation are still manually performed, which is time-consuming and prone to errors.
1 code implementation • 29 Oct 2018 • Arnab Kumar Mondal, Jose Dolz, Christian Desrosiers
In addition, our work presents a comprehensive analysis of different GAN architectures for semi-supervised segmentation, showing recent techniques like feature matching to yield a higher performance than conventional adversarial training approaches.
no code implementations • 16 Oct 2018 • Jose Dolz, Ismail Ben Ayed, Christian Desrosiers
First, instead of combining the available image modalities at the input, each of them is processed in a different path to better exploit their unique information.
1 code implementation • 9 Jul 2018 • Aarush Gupta, Dakshit Agrawal, Hardik Chauhan, Jose Dolz, Marco Pedersoli
In this paper we propose a new approach for classifying the global emotion of images containing groups of people.
no code implementations • 28 May 2018 • Jose Dolz, Xiaopan Xu, Jerome Rony, Jing Yuan, Yang Liu, Eric Granger, Christian Desrosiers, Xi Zhang, Ismail Ben Ayed, Hongbing Lu
Precise segmentation of bladder walls and tumor regions is an essential step towards non-invasive identification of tumor stage and grade, which is critical for treatment decision and prognosis of patients with bladder cancer (BC).
4 code implementations • 12 May 2018 • Hoel Kervadec, Jose Dolz, Meng Tang, Eric Granger, Yuri Boykov, Ismail Ben Ayed
To the best of our knowledge, the method of [Pathak et al., 2015] is the only prior work that addresses deep CNNs with linear constraints in weakly supervised segmentation.
3 code implementations • 9 Apr 2018 • Jose Dolz, Karthik Gopinath, Jing Yuan, Herve Lombaert, Christian Desrosiers, Ismail Ben Ayed
Therefore, the proposed network has total freedom to learn more complex combinations between the modalities, within and in-between all the levels of abstraction, which increases significantly the learning representation.
Ranked #1 on Medical Image Segmentation on iSEG 2017 Challenge
1 code implementation • 14 Dec 2017 • Jose Dolz, Christian Desrosiers, Li Wang, Jing Yuan, Dinggang Shen, Ismail Ben Ayed
We report evaluations of our method on the public data of the MICCAI iSEG-2017 Challenge on 6-month infant brain MRI segmentation, and show very competitive results among 21 teams, ranking first or second in most metrics.
Ranked #1 on Infant Brain Mri Segmentation on iSEG 2017 Challenge
1 code implementation • 16 Oct 2017 • Jose Dolz, Ismail Ben Ayed, Jing Yuan, Christian Desrosiers
Neonatal brain segmentation in magnetic resonance (MR) is a challenging problem due to poor image quality and low contrast between white and gray matter regions.
1 code implementation • 28 Apr 2017 • Jose Dolz, Ismail Ben Ayed, Christian Desrosiers
We propose to constrain segmentation functionals with a dimensionless, unbiased and position-independent shape compactness prior, which we solve efficiently with an alternating direction method of multipliers (ADMM).
no code implementations • 21 Apr 2017 • Tobias Fechter, Sonja Adebahr, Dimos Baltas, Ismail Ben Ayed, Christian Desrosiers, Jose Dolz
These figures translate into a very good agreement with the reference contours and an increase in accuracy compared to other methods.
no code implementations • CVPR 2017 • Jose Dolz, Ismail Ben Ayed, Christian Desrosiers
We formulate an Alternating Direction Method of Mul-tipliers (ADMM) that systematically distributes the computations of any technique for optimizing pairwise functions, including non-submodular potentials.
no code implementations • 30 Mar 2017 • Jose Dolz, Nicolas Reyns, Nacim Betrouni, Dris Kharroubi, Mathilde Quidet, Laurent Massoptier, Maximilien Vermandel
Prescribed radiation therapy for brain cancer requires precisely defining the target treatment area, as well as delineating vital brain structures which must be spared from radiotoxicity.