Etienne Meunier
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Etienne Meunier has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Patrick Bouthemy
INRIA
2 shared publications Anais Badoual
INRIA, Centre Rennes - Bretagne Atlantique, Rennes Cedex, France
1 shared publication- David Kamm
Sorbonne Université-CNRS-IRD-MNHN, LOCEAN Laboratory, Paris, France
1 shared publication - Eugene Pinsky
MET Department of Computer Science, Boston University, Boston, MA 02215, USA
1 shared publication - Félicien Catherin
Félicien Catherin EI, Chatillon, France
1 shared publication Guillaume Gachon
1 shared publication
Latest Publications
- 9 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 48 (1) , 500-511
Segmenting the Motion Components of a Video: A Long-Term Unsupervised Model
Human beings have the ability to continuously analyze a video and immediately extract the motion components. We want to adopt this paradigm to provide a coherent and stable motion segmentation over the video sequence. In this perspective, we propose a novel long-term spatio-temporal model operating in a totally unsupervised way. It takes as input the volume of consecutive optical flow (OF) fields, and delivers a volume of segments of coherent motion over the video. More specifically, we have designed a transformer-based network, where we leverage a mathematically well-founded framework, the Evidence Lower Bound (ELBO), to derive the loss function. The loss function combines a flow reconstruction term involving spatio-temporal parametric motion models combining, in a novel way, polynomial (quadratic) motion models for the spatial dimensions and B-splines for the time dimension of the video sequence, and a regularization term enforcing temporal consistency on the segments. We report experiments on four VOS benchmarks, demonstrating competitive quantitative results while performing motion segmentation on a sequence in one go. We also highlight through visual results the key contributions on temporal consistency brought by our method.
- 9 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 48 (1) , 500-511
Segmenting the Motion Components of a Video: A Long-Term Unsupervised Model
Human beings have the ability to continuously analyze a video and immediately extract the motion components. We want to adopt this paradigm to provide a coherent and stable motion segmentation over the video sequence. In this perspective, we propose a novel long-term spatio-temporal model operating in a totally unsupervised way. It takes as input the volume of consecutive optical flow (OF) fields, and delivers a volume of segments of coherent motion over the video. More specifically, we have designed a transformer-based network, where we leverage a mathematically well-founded framework, the Evidence Lower Bound (ELBO), to derive the loss function. The loss function combines a flow reconstruction term involving spatio-temporal parametric motion models combining, in a novel way, polynomial (quadratic) motion models for the spatial dimensions and B-splines for the time dimension of the video sequence, and a regularization term enforcing temporal consistency on the segments. We report experiments on four VOS benchmarks, demonstrating competitive quantitative results while performing motion segmentation on a sequence in one go. We also highlight through visual results the key contributions on temporal consistency brought by our method.
- Open Access
- 18 March 2025
- Published by Copernicus GmbH
Can diffusion model generate ocean states compatible with OGCM ?
Ocean General Circulation Models (hereafter OGCM) are critical to the study of past and present climate, and the production of future projections. Unfortunately, they require large amounts of computations at simulation time. On the other hand, deep learning emulators trained on reanalyses are starting to deliver accurate short-term predictions, using comparatively small computational resources, yet they struggle to deliver long term predictions, are not interpretable and do no't factor in the uncertainty in physical parameters. As a result, they cannot be used by climate scientists to understand mechanisms of climate variability, such as tipping points, nor adjustment processes to greenhouse gas emissions.Aiming to take the best from each world and establish a close interaction between emulators and OGCM, we investigate whether an emulator can be used to provide state variables to an ocean model, which would then handle the temporal integration using physics equations. Namely, we trained a diffusion model on a large dataset of ocean variables produced by NEMO, analysed the newly generated states, propose metrics to assess their physical consistency, and use them as initial conditions of simulations to assess their compatibility (physical and numerical) with NEMO.Overall, we want to determine whether unconstrained generative models are able to produce realistic solutions, and to assess the tolerance of OGCM to externally generated ocean states, what we consider as a first step towards building an hybrid OGCM.
- 18 March 2025
- Published by Copernicus GmbH
Open AccessCan diffusion model generate ocean states compatible with OGCM ?
Ocean General Circulation Models (hereafter OGCM) are critical to the study of past and present climate, and the production of future projections. Unfortunately, they require large amounts of computations at simulation time. On the other hand, deep learning emulators trained on reanalyses are starting to deliver accurate short-term predictions, using comparatively small computational resources, yet they struggle to deliver long term predictions, are not interpretable and do no't factor in the uncertainty in physical parameters. As a result, they cannot be used by climate scientists to understand mechanisms of climate variability, such as tipping points, nor adjustment processes to greenhouse gas emissions.Aiming to take the best from each world and establish a close interaction between emulators and OGCM, we investigate whether an emulator can be used to provide state variables to an ocean model, which would then handle the temporal integration using physics equations. Namely, we trained a diffusion model on a large dataset of ocean variables produced by NEMO, analysed the newly generated states, propose metrics to assess their physical consistency, and use them as initial conditions of simulations to assess their compatibility (physical and numerical) with NEMO.Overall, we want to determine whether unconstrained generative models are able to produce realistic solutions, and to assess the tolerance of OGCM to externally generated ocean states, what we consider as a first step towards building an hybrid OGCM.
- Open Access
- 19 December 2023
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
- Vol. 7 (4) , 1-25
ADA-SHARK
Due to global warming, sharks are moving closer to the beaches, affecting the risk to humans and their own lives. Within the past decade, several technologies were developed to reduce the risks for swimmers and surfers. This study proposes a robust method based on computer vision to detect sharks using an underwater camera monitoring system to secure coastlines. The system is autonomous, environment-friendly, and requires low maintenance. 43,679 images extracted from 175 hours of videos of marine life were used to train our algorithms. Our approach allows the collection and analysis of videos in real-time using an autonomous underwater camera connected to a smart buoy charged with solar panels. The videos are processed by a Domain Adversarial Convolutional Neural Network to discern sharks regardless of the background environment with an F2-score of 83.2% and a recall of 90.9%, while human experts have an F2-score of 94% and a recall of 95.7%.
- 19 December 2023
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
- Vol. 7 (4) , 1-25
Open AccessADA-SHARK
Due to global warming, sharks are moving closer to the beaches, affecting the risk to humans and their own lives. Within the past decade, several technologies were developed to reduce the risks for swimmers and surfers. This study proposes a robust method based on computer vision to detect sharks using an underwater camera monitoring system to secure coastlines. The system is autonomous, environment-friendly, and requires low maintenance. 43,679 images extracted from 175 hours of videos of marine life were used to train our algorithms. Our approach allows the collection and analysis of videos in real-time using an autonomous underwater camera connected to a smart buoy charged with solar panels. The videos are processed by a Domain Adversarial Convolutional Neural Network to discern sharks regardless of the background environment with an F2-score of 83.2% and a recall of 90.9%, while human experts have an F2-score of 94% and a recall of 95.7%.
- 15 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. PP (01628828) , 1-12
EM-Driven Unsupervised Learning for Efficient Motion Segmentation
In this paper, we present a CNN-based fully unsupervised method for motion segmentation from optical flow. We assume that the input optical flow can be represented as a piecewise set of parametric motion models, typically, affine or quadratic motion models. The core idea of our work is to leverage the Expectation-Maximization (EM) framework in order to design in a well-founded manner a loss function and a training procedure of our motion segmentation neural network that does not require either ground-truth or manual annotation. However, in contrast to the classical iterative EM, once the network is trained, we can provide a segmentation for any unseen optical flow field in a single inference step and without estimating any motion models. We investigate different loss functions including robust ones and propose a novel efficient data augmentation technique on the optical flow field, applicable to any network taking optical flow as input. In addition, our method is able by design to segment multiple motions. Our motion segmentation network was tested on four benchmarks, DAVIS2016, SegTrackV2, FBMS59, and MoCA, and performed very well, while being fast at test time.
- 15 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. PP (01628828) , 1-12
EM-Driven Unsupervised Learning for Efficient Motion Segmentation
In this paper, we present a CNN-based fully unsupervised method for motion segmentation from optical flow. We assume that the input optical flow can be represented as a piecewise set of parametric motion models, typically, affine or quadratic motion models. The core idea of our work is to leverage the Expectation-Maximization (EM) framework in order to design in a well-founded manner a loss function and a training procedure of our motion segmentation neural network that does not require either ground-truth or manual annotation. However, in contrast to the classical iterative EM, once the network is trained, we can provide a segmentation for any unseen optical flow field in a single inference step and without estimating any motion models. We investigate different loss functions including robust ones and propose a novel efficient data augmentation technique on the optical flow field, applicable to any network taking optical flow as input. In addition, our method is able by design to segment multiple motions. Our motion segmentation network was tested on four benchmarks, DAVIS2016, SegTrackV2, FBMS59, and MoCA, and performed very well, while being fast at test time.