Showing 1–5 of 5 results for author: Khalfaoui, S

DiffHPE: Robust, Coherent 3D Human Pose Lifting with Diffusion

Authors: Cédric Rommel, Eduardo Valle, Mickaël Chen, Souhaiel Khalfaoui, Renaud Marlet, Matthieu Cord, Patrick Pérez

Abstract: We present an innovative approach to 3D Human Pose Estimation (3D-HPE) by integrating cutting-edge diffusion models, which have revolutionized diverse fields, but are relatively unexplored in 3D-HPE. We show that diffusion models enhance the accuracy, robustness, and coherence of human pose estimations. We introduce DiffHPE, a novel strategy for harnessing diffusion models in 3D-HPE, and demonstra… ▽ More We present an innovative approach to 3D Human Pose Estimation (3D-HPE) by integrating cutting-edge diffusion models, which have revolutionized diverse fields, but are relatively unexplored in 3D-HPE. We show that diffusion models enhance the accuracy, robustness, and coherence of human pose estimations. We introduce DiffHPE, a novel strategy for harnessing diffusion models in 3D-HPE, and demonstrate its ability to refine standard supervised 3D-HPE. We also show how diffusion models lead to more robust estimations in the face of occlusions, and improve the time-coherence and the sagittal symmetry of predictions. Using the Human\,3.6M dataset, we illustrate the effectiveness of our approach and its superiority over existing models, even under adverse situations where the occlusion patterns in training do not match those in inference. Our findings indicate that while standalone diffusion models provide commendable performance, their accuracy is even better in combination with supervised models, opening exciting new avenues for 3D-HPE research. △ Less

Submitted 4 September, 2023; originally announced September 2023.

Comments: Accepted to 2023 International Conference on Computer Vision Workshop (Analysis and Modeling of Faces and Gestures)

Goppa-like AG codes from $C_{a,b}$ curves and their behaviour under squaring their dual

Authors: Sabira El Khalfaoui, Mathieu Lhotel, Jade Nardi

Abstract: In this paper, we introduce a family of codes that can be used in a McEliece cryptosystem, called Goppa--like AG codes. These codes generalize classical Goppa codes and can be constructed from any curve of genus $\mathfrak{g} \geq 0$. Focusing on codes from $C_{a,b}$ curves, we study the behaviour of the dimension of the square of their dual to determine their resistance to distinguisher attacks s… ▽ More In this paper, we introduce a family of codes that can be used in a McEliece cryptosystem, called Goppa--like AG codes. These codes generalize classical Goppa codes and can be constructed from any curve of genus $\mathfrak{g} \geq 0$. Focusing on codes from $C_{a,b}$ curves, we study the behaviour of the dimension of the square of their dual to determine their resistance to distinguisher attacks similar to the one for alternant and Goppa codes developed by Mora and Tillich. We also propose numerical experiments to measure how sharp is our bound. △ Less

Submitted 14 April, 2023; v1 submitted 15 March, 2023; originally announced March 2023.

Comments: Minor changes: authors reordered alphabetically and missing parentheses added in Corollary 1.8

MSC Class: 11T71; 14G50; 14H05; 11G20

Empowering the trustworthiness of ML-based critical systems through engineering activities

Authors: Juliette Mattioli, Agnes Delaborde, Souhaiel Khalfaoui, Freddy Lecue, Henri Sohier, Frederic Jurie

Abstract: This paper reviews the entire engineering process of trustworthy Machine Learning (ML) algorithms designed to equip critical systems with advanced analytics and decision functions. We start from the fundamental principles of ML and describe the core elements conditioning its trust, particularly through its design: namely domain specification, data engineering, design of the ML algorithms, their im… ▽ More This paper reviews the entire engineering process of trustworthy Machine Learning (ML) algorithms designed to equip critical systems with advanced analytics and decision functions. We start from the fundamental principles of ML and describe the core elements conditioning its trust, particularly through its design: namely domain specification, data engineering, design of the ML algorithms, their implementation, evaluation and deployment. The latter components are organized in an unique framework for the design of trusted ML systems. △ Less

Submitted 30 September, 2022; originally announced September 2022.

Comments: This work has been supported by the French government under the "France 2030" program, as part of the SystemX Technological Research Institute Research Institute

Estimating The Dimension Of The Subfield Subcodes of Hermitian Codes

Authors: Sabira El Khalfaoui, Gábor P. Nagy

Abstract: In this paper, we study the behavior of the true dimension of the subfield subcodes of Hermitian codes. Our motivation is to use these classes of linear codes to improve the parameters of the McEliece cryptosystem, such that key size and security level. The McEliece scheme is one of the promising alternative cryptographic schemes to the current public key schemes since in the last four decades, th… ▽ More In this paper, we study the behavior of the true dimension of the subfield subcodes of Hermitian codes. Our motivation is to use these classes of linear codes to improve the parameters of the McEliece cryptosystem, such that key size and security level. The McEliece scheme is one of the promising alternative cryptographic schemes to the current public key schemes since in the last four decades, they resisted all known quantum computing attacks. By analyzing computational data series of true dimension, we concluded that they can be estimated by the extreme value distribution function. △ Less

Submitted 1 April, 2020; originally announced April 2020.

MSC Class: 11T71; 14G50; 94B27

On The Dimension of The Subfield Subcodes of 1-Point Hermitian Codes

Authors: Sabira El Khalfaoui, Gábor P. Nagy

Abstract: Subfield subcodes of algebraic-geometric codes are good candidates for the use in post-quantum cryptosystems, provided their true parameters such as dimension and minimum distance can be determined. In this paper we present new values of the true dimension of subfield subcodes of $1$--point Hermitian codes, including the case when the subfield is not binary. Subfield subcodes of algebraic-geometric codes are good candidates for the use in post-quantum cryptosystems, provided their true parameters such as dimension and minimum distance can be determined. In this paper we present new values of the true dimension of subfield subcodes of $1$--point Hermitian codes, including the case when the subfield is not binary. △ Less

Submitted 9 September, 2019; v1 submitted 25 June, 2019; originally announced June 2019.

MSC Class: 11T71; 14G50; 94B27