Erwann Kervennic
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Erwann Kervennic has a total of 12 co-authors. Here are the top co-authors by number of shared publications:
- Ingrid Grenet
Institut de Recherche Technologique Saint Exupry, Biot, France
2 shared publications Adrien Girard
IRT,Saint Exupry,France
1 shared publication- Benjamin Francesconi
Institut de Recherche Technologique Saint Exupry, Biot, France
1 shared publication Frédéric Férésin
IRT,Saint Exupry,France
1 shared publication- Marjorie Bellizzi
Institut de Recherche Technologique Saint Exupry, Biot, France
1 shared publication Michael Benguigui
IRT,Saint Exupry,France
1 shared publication
Latest Publications
- Open Access
- 27 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 17 (19391404) , 7918-7931
Onboard Anomaly Detection for Marine Environmental Protection
This article describes a comprehensive artificial intelligence pipeline for detecting threatening events in the marine environment, which is intended to be run on-board Earth observation satellites and globally contribute to the preservation of the marine environment. We employed a self-supervised neural network-based anomaly detection technique to identify a wide range of potentially unknown events (pollutions). This method consist in identifying deviations from the learned “normal” water state as abnormal events. Then, we can better characterize some of the anomalies by incorporating other, more selective models into our processing pipeline to prioritize targeted actions over specific anomalies (e.g., oil spills). This approach's effectiveness and versatility are demonstrated through its selection in two European Space Agency competitive challenges aimed at integrating artificial intelligence into two Earth observation demonstration missions: $\Phi$ sat-2 and IMAGIN-e. The article details the processing pipeline design, the generated datasets for training and testing steps of the two challenges, and the detection effectiveness and hardware efficiency (including throughput and latency) when embedded on missions-representative hardware. Our primary contribution is the proposal of an operational pipeline with the constraints to (1) detect multiple and unknown events, (2) require only “weakly” labelled data and (3) be computationally efficient enough to run on low-power targets on-board Earth observation payloads. These three assets make this pipeline easily adaptable to a wide range of missions.
- 27 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 17 (19391404) , 7918-7931
Open AccessOnboard Anomaly Detection for Marine Environmental Protection
This article describes a comprehensive artificial intelligence pipeline for detecting threatening events in the marine environment, which is intended to be run on-board Earth observation satellites and globally contribute to the preservation of the marine environment. We employed a self-supervised neural network-based anomaly detection technique to identify a wide range of potentially unknown events (pollutions). This method consist in identifying deviations from the learned “normal” water state as abnormal events. Then, we can better characterize some of the anomalies by incorporating other, more selective models into our processing pipeline to prioritize targeted actions over specific anomalies (e.g., oil spills). This approach's effectiveness and versatility are demonstrated through its selection in two European Space Agency competitive challenges aimed at integrating artificial intelligence into two Earth observation demonstration missions: $\Phi$ sat-2 and IMAGIN-e. The article details the processing pipeline design, the generated datasets for training and testing steps of the two challenges, and the detection effectiveness and hardware efficiency (including throughput and latency) when embedded on missions-representative hardware. Our primary contribution is the proposal of an operational pipeline with the constraints to (1) detect multiple and unknown events, (2) require only “weakly” labelled data and (3) be computationally efficient enough to run on low-power targets on-board Earth observation payloads. These three assets make this pipeline easily adaptable to a wide range of missions.
- 2 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Embedded cloud segmentation using AI : Back on years of experiments in orbit on OPS-SAT
Since 2019, IRT Saint Exupéry has been researching embedded cloud segmentation and has conducted several experiments on board the OPS-SAT satellite from the European Space Agency. Using FPGA implementations that efficiently execute artificial neural networks, an image with dimensions 2048×1944×3 is inferred in less than 126 ms while consuming less than 2W of power. These neural network inferences in the programmable logic part of a FPGA are, to our knowledge, a first in orbit.In this paper, we summarize the work and main results obtained by IRT Saint Exupéry during the whole OPS-SAT mission. We start by describing and comparing our main neural network topologies for cloud segmentation together with a ZGP formula, an ultra-light mathematical equation. Keeping only the best model, we then train it on various evolutions of databases built over the years. Deploying this network on OPS-SAT and on a set of other hardware targets of interest (Google Coral / Intel Neural Compute Stick 2), we finally demonstrate that the processing throughput on FPGA is 10 to 36 times faster than on manufacturer-specific ASICs with an equivalent power consumption and better overall algorithmic performance.This paper also contains original material. In particular, it details the construction of a generic VHDL library developed by IRT Saint Exupéry, and the other tools and methods we used in the context of the CIAR project.
- 2 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Embedded cloud segmentation using AI : Back on years of experiments in orbit on OPS-SAT
Since 2019, IRT Saint Exupéry has been researching embedded cloud segmentation and has conducted several experiments on board the OPS-SAT satellite from the European Space Agency. Using FPGA implementations that efficiently execute artificial neural networks, an image with dimensions 2048×1944×3 is inferred in less than 126 ms while consuming less than 2W of power. These neural network inferences in the programmable logic part of a FPGA are, to our knowledge, a first in orbit.In this paper, we summarize the work and main results obtained by IRT Saint Exupéry during the whole OPS-SAT mission. We start by describing and comparing our main neural network topologies for cloud segmentation together with a ZGP formula, an ultra-light mathematical equation. Keeping only the best model, we then train it on various evolutions of databases built over the years. Deploying this network on OPS-SAT and on a set of other hardware targets of interest (Google Coral / Intel Neural Compute Stick 2), we finally demonstrate that the processing throughput on FPGA is 10 to 36 times faster than on manufacturer-specific ASICs with an equivalent power consumption and better overall algorithmic performance.This paper also contains original material. In particular, it details the construction of a generic VHDL library developed by IRT Saint Exupéry, and the other tools and methods we used in the context of the CIAR project.