Antitza Dantcheva
Portrait of Antitza Dantcheva

Antitza Dantcheva

Research Director (directrice de recherche), STARS team, Inria Centre at Université Côte d'Azur
3IA Côte d'Azur Chair

A human face reveals traits such as the identity, expression, age and intent. My research aims at discovering how machines learn, generate and separate what a face encodes.

Most recently, my group has been uncovering the structure that generative models learn about human faces, and using it to steer them without retraining. We also design models that animate portraits and generate human video, and methods for deepfake detection, privacy and health.

News

  • Accepted at NeurIPS 2026: N. Quignon, A. Dantcheva. Few-shot visual concept extraction for steering diffusion transformers.
  • Accepted at WACV 2027: C. Yahchouchi*, N. Quignon*, N. Roggero, L. Saroul, A. Dantcheva (*equal contribution). MoSTE: mixture of spatio-temporal experts for deepfake detection in videos.
  • I am Program Chair of ACM Multimedia 2026 (Rio de Janeiro, 10 to 14 November 2026).
  • I am Program Chair of IEEE FG 2027 (Marrakesh, 26 to 30 April 2027). Our special session "Artificial Intelligence for Automated Human Health-care and Monitoring," with Abhijit Das and François Brémond, was accepted for FG 2027.
  • Oral at the ABAW workshop of ECCV 2026: B. Pokrzywa, N. Quignon, Y. Bahram, M. O. Zeeshan, A. Dantcheva, E. Granger. BEACON: behavior and appearance control for subject-specific video generation.
  • Published in IJCV: B. Chopin, T. Dhamija, P. Balaji, Y. Wang, A. Dantcheva. AI killed the video star: audio-driven diffusion model for expressive talking head generation.
  • Published in IEEE T-BIOM: S. Atreya, N. Quignon, B. Chopin, A. Das, A. Dantcheva. Beyond real versus fake: towards intent-aware video analysis.
  • Published in IEEE T-BIOM: M. Bora, T. Dhamija, S. Reddy, B. Chopin, P. Balaji, A. Das, A. Dantcheva. Do you see what I say? Generalizable deepfake detection based on visual speech recognition.
  • In the CVPR 2026 Findings track: N. Quignon, B. Chopin, Y. Wang, A. Dantcheva. THEval: evaluation framework for talking head video generation.
  • Co-organizer of the 5th RePSS challenge and workshop on remote physiological signal sensing, at IJCAI 2026 in Bremen.
  • I was Program Chair of ICIG 2026.
  • Dr. Snehashis Majhi defended his Ph.D. and joined Woven by Toyota as a Research Scientist.
  • Invited talk at Roma Tre University, Rome.
  • I was appointed 3IA Côte d'Azur Chair in the Core AI axis, on "Generating synthetic videos for automated visual learning."
Show older news
  • Co-organizer of CV4BIOM, the Workshop on Computer Vision for Biometrics, Identity and Behaviour, at ICCV 2025, and of the 4th RePSS workshop at IJCAI 2025.
  • Dr. Valeriya Strizhkova (March) and Dr. Tanay Agrawal defended their Ph.D.s; Tanay received the Université Côte d'Azur Ph.D. Prize. Both have since co-founded start-ups, Facila and AreWeMe.
  • Invited talks at the SRBS workshop at BMVC, TU Wien, the SophIA Summit and the Deepfakes Workshop of the US Strategy Forum, New York.
  • At CVPR 2025: S. Majhi, G. D'Amicantonio, A. Dantcheva, Q. Kong, L. Garattoni, G. Francesca, E. Bondarev, F. Brémond. Just dance with π! A poly-modal inductor for weakly-supervised video anomaly detection.
  • On arXiv: Y. Wang, D. Yang, X. Chen, F. Brémond, Y. Qiao, A. Dantcheva. LIA-X: interpretable latent portrait animator.
  • Promoted to directrice de recherche at Inria. Visit and talk at BITS Pilani, Hyderabad, within our associate team on generalizable deepfake detection.
  • Top-3 finalist, Outstanding Women in Technology South of France. Dr. David Anghelone received the Thales Ph.D. Prize.
  • Talk on deciphering human faces with AI at the centenary of the "illusion des sosies" (Capgras syndrome), Société Médico-Psychologique, Paris. Lecture on deepfakes at the spring school on multimodal foundation models and generative AI, Rabat. Talk at the University of Cagliari.
  • Dr. David Anghelone defended his Ph.D. on face recognition in the invisible spectrum and received the Université Côte d'Azur Ph.D. Prize. Tutorial "Alpha and omega of deepfakes," with Abhijit Das, at IJCB 2023; invited talk at INSAIT, Sofia.
  • My first Ph.D. student, Dr. Yaohui Wang, defended his Ph.D. "Learning to generate human videos," and I defended my habilitation "Computer vision for deciphering and generating faces."

Research

What generative models learn about faces

Diffusion and video models now synthesise faces with extraordinary fidelity, yet what they have learned about people remains largely unexamined. We isolate human visual concepts, such as a smile, age or a beard, as directions inside frozen models, study how these directions are organised, and compare that organisation with what psychology has established about face perception.

  • N. Quignon, A. Dantcheva. Few-shot visual concept extraction for steering diffusion transformers. NeurIPS 2026.
  • B. Pokrzywa, N. Quignon, Y. Bahram, M. O. Zeeshan, A. Dantcheva, E. Granger. BEACON: behavior and appearance control for subject-specific video generation. ECCV 2026 ABAW Workshop (oral). arXiv
Visual concepts extracted from a few examples steer a frozen diffusion transformer.
Steering a frozen diffusion transformer with visual concepts extracted from a few examples (N. Quignon, A. Dantcheva, NeurIPS 2026).

Generating human video

We animate still portraits and generate human video by navigating interpretable directions in a model's latent space, instead of relying on facial keypoints or 3D face models, and we develop ways to evaluate the result.

  • Y. Wang, P. Bilinski, F. Brémond, A. Dantcheva. G3AN: disentangling appearance and motion for video generation. CVPR 2020. arXiv
  • Y. Wang, D. Yang, F. Brémond, A. Dantcheva. LIA: Latent Image Animator. IEEE TPAMI 2024, extending ICLR 2022. arXivcode
  • Y. Wang, X. Ma, X. Chen, C. Chen, A. Dantcheva, B. Dai, Y. Qiao. LEO: generative latent image animator for human video synthesis. IJCV 2024. arXivcode
  • Y. Wang, D. Yang, X. Chen, F. Brémond, Y. Qiao, A. Dantcheva. LIA-X: interpretable latent portrait animator. arXiv 2025. arXivproject
  • B. Chopin, T. Dhamija, P. Balaji, Y. Wang, A. Dantcheva. AI killed the video star: audio-driven diffusion model for expressive talking head generation. IJCV 2026. arXiv
  • N. Quignon, B. Chopin, Y. Wang, A. Dantcheva. THEval: evaluation framework for talking head video generation. CVPR 2026 Findings. arXivproject
Overview of LIA-X, an interpretable latent portrait animator.
LIA-X, our interpretable latent portrait animator (Y. Wang, D. Yang, X. Chen, F. Brémond, Y. Qiao, A. Dantcheva, 2025).

Deepfake detection and trustworthy face analysis

We detect manipulated video and ask what a manipulation intends, and we study what faces reveal and how to keep that knowledge fair and safe: removing identity while preserving expression, estimating bias, and detecting apathy from facial behaviour.

  • C. Yahchouchi*, N. Quignon*, N. Roggero, L. Saroul, A. Dantcheva (*equal contribution). MoSTE: mixture of spatio-temporal experts for deepfake detection in videos. WACV 2027.
  • S. Atreya, N. Quignon, B. Chopin, A. Das, A. Dantcheva. Beyond real versus fake: towards intent-aware video analysis. IEEE T-BIOM 2026. arXiv
  • M. Bora, T. Dhamija, S. Reddy, B. Chopin, P. Balaji, A. Das, A. Dantcheva. Do you see what I say? Generalizable deepfake detection based on visual speech recognition. IEEE T-BIOM 2026. arXiv
  • A. Das, X. Niu, A. Dantcheva, S. L. Happy, H. Han, R. Zeghari, P. Robert, S. Shan, F. Brémond, X. Chen. A spatio-temporal approach for apathy classification. IEEE TCSVT 2022.
Architecture of MoSTE: a CLIP ViT backbone with MoSTE adapters, in which spatial and temporal routers send video tokens to spatial and temporal attention experts.
MoSTE, our mixture of spatio-temporal experts for deepfake detection (C. Yahchouchi, N. Quignon, N. Roggero, L. Saroul, A. Dantcheva, WACV 2027).

Selected publications

2027

  • C. Yahchouchi*, N. Quignon*, N. Roggero, L. Saroul, A. Dantcheva (*equal contribution). MoSTE: mixture of spatio-temporal experts for deepfake detection in videos. WACV 2027.

2026

  • N. Quignon, A. Dantcheva. Few-shot visual concept extraction for steering diffusion transformers. NeurIPS 2026.
  • B. Chopin, T. Dhamija, P. Balaji, Y. Wang, A. Dantcheva. AI killed the video star: audio-driven diffusion model for expressive talking head generation. IJCV 2026. arXiv
  • S. Atreya, N. Quignon, B. Chopin, A. Das, A. Dantcheva. Beyond real versus fake: towards intent-aware video analysis. IEEE T-BIOM 2026. arXiv
  • M. Bora, T. Dhamija, S. Reddy, B. Chopin, P. Balaji, A. Das, A. Dantcheva. Do you see what I say? Generalizable deepfake detection based on visual speech recognition. IEEE T-BIOM 2026. arXiv
  • N. Quignon, B. Chopin, Y. Wang, A. Dantcheva. THEval: evaluation framework for talking head video generation. CVPR 2026 Findings. arXivproject
  • B. Pokrzywa, N. Quignon, Y. Bahram, M. O. Zeeshan, A. Dantcheva, E. Granger. BEACON: behavior and appearance control for subject-specific video generation. ECCV 2026 ABAW Workshop (oral). arXiv

2025

  • S. Majhi, G. D'Amicantonio, A. Dantcheva, Q. Kong, L. Garattoni, G. Francesca, E. Bondarev, F. Brémond. Just dance with π! A poly-modal inductor for weakly-supervised video anomaly detection. CVPR 2025. arXivcode
  • Y. Wang, D. Yang, X. Chen, F. Brémond, Y. Qiao, A. Dantcheva. LIA-X: interpretable latent portrait animator. arXiv 2025. arXivproject

2024 and earlier

  • Y. Wang, D. Yang, F. Brémond, A. Dantcheva. LIA: Latent Image Animator. IEEE TPAMI 2024, extending ICLR 2022. arXivcode
  • Y. Wang, X. Ma, X. Chen, C. Chen, A. Dantcheva, B. Dai, Y. Qiao. LEO: generative latent image animator for human video synthesis. IJCV 2024. arXivcode
  • I. Joshi, M. Grimmer, C. Rathgeb, C. Busch, F. Brémond, A. Dantcheva. Synthetic data in human analysis: a survey. IEEE TPAMI 2024. arXiv
  • H. Chen, Y. Wang, B. Lagadec, A. Dantcheva, F. Brémond. Learning invariance from generated variance for unsupervised person re-identification. IEEE TPAMI 2023. arXiv
  • A. Das, X. Niu, A. Dantcheva, S. L. Happy, H. Han, R. Zeghari, P. Robert, S. Shan, F. Brémond, X. Chen. A spatio-temporal approach for apathy classification. IEEE TCSVT 2022.
  • Y. Wang, P. Bilinski, F. Brémond, A. Dantcheva. G3AN: disentangling appearance and motion for video generation. CVPR 2020. arXiv
  • A. Dantcheva, P. Elia, A. Ross. What else does your biometric data reveal? A survey on soft biometrics. IEEE TIFS 2016. PDF
  • A. Dantcheva, C. Chen, A. Ross. Can facial cosmetics affect the matching accuracy of face recognition systems? IEEE BTAS 2012.

The full list is on Google Scholar. Datasets: makeup datasets, VIP attribute dataset.

Team

Current members

  • Nabyl Quignon, Ph.D. student, concept discovery and steering in generative models
  • Charbel Yahchouchi, Ph.D. student (CIFRE with Probayes), deepfake detection
  • Mahmoud Ali, Ph.D. student co-supervised with François Brémond, human-centered video understanding
  • Felix Kroiss, research intern, steering of diffusion transformers

Ph.D. alumni

  • Dr. Snehashis Majhi (2026), Research Scientist, Woven by Toyota
  • Dr. Valeriya Strizhkova (2025), co-founder, Facila
  • Dr. Tanay Agrawal (2025, Université Côte d'Azur Ph.D. Prize), co-founder, AreWeMe
  • Dr. Arnab Dey (2024), Senior Research Engineer, UNITH, Barcelona
  • Dr. David Anghelone (2023, Thales Ph.D. Prize), AI Specialist, Groupe Alaine
  • Dr. Yaohui Wang (2021), Head of the Vorch video generation team, Bilibili
  • Dr. Ester Gonzalez-Sosa (2017, Universidad Autónoma de Madrid), Machine Learning Research Engineer, Nokia
  • Dr. Cunjian Chen (2014, West Virginia University), Associate Professor, Monash University

Former postdoctoral fellows

  • Dr. Baptiste Chopin, Research Associate, Hochschule Darmstadt
  • Dr. Indu Joshi, Assistant Professor, IIT Mandi
  • Dr. Michal Balazia, permanent researcher, Inria
  • Dr. Abhijit Das, Assistant Professor, BITS Pilani
  • Dr. S. L. Happy, Data Scientist, HP

Service and recognition

Service

  • 2027Program Chair, IEEE FG 2027
  • 2026Program Chair, ACM Multimedia 2026 and ICIG 2026
  • 2026Co-organizer, 5th RePSS challenge and workshop, IJCAI 2026
  • 2025Co-organizer, CV4BIOM workshop at ICCV 2025 and 4th RePSS workshop at IJCAI 2025
  • 2024Workshop Chair, CVPR 2024; Area Chair, ECCV 2024
  • 2023Tutorial Chair, IJCB 2023
  • 2022Area Chair, ACM Multimedia 2022; consulted by the French Senate on facial recognition
  • 2019Publication Chair, IEEE FG 2019
  • 2017Program Chair, BIOSIG (2017 to 2020)

Associate Editor of Pattern Recognition, IEEE Transactions on Multimedia and Multimedia Tools and Applications. Member of ELLIS and of the European Association for Biometrics.

Awards and fellowships

  • 20253IA Côte d'Azur Chair
  • 2025Université Côte d'Azur Ph.D. Prize for my student Tanay Agrawal
  • 2024Top-3 finalist, Outstanding Women in Technology South of France
  • 2024Thales Ph.D. Prize for my student David Anghelone (Université Côte d'Azur Ph.D. Prize, 2023)
  • 2022Winner, New Technologies Show, ECCV 2022 (start-up project Verbalia)
  • 2019Best Poster Award, IEEE FG 2019
  • 2018Winner, Bias Estimation in Face Analytics (BEFA) challenge, ECCV 2018
  • 2017Best Paper Award (runner-up), IEEE ISBA 2017
  • 2016Labex postdoctoral fellowship, Université Côte d'Azur
  • 2014ERCIM Alain Bensoussan (Marie Curie) fellowship
  • 2013Best Spoofing Attack Award, Tabula Rasa competition; Best Poster Award, IAPR ICB 2013
  • 2011IBM Best Presentation Award, IEEE ICME 2011

Interviews in Le Monde and Science & Vie; talk "Generating and deciphering faces".

Open positions

I am looking for Ph.D. students, postdocs and interns interested in deepfake detection: detecting manipulated and generated human video, understanding what a manipulation intends, and making detectors that generalise to new generators.

Send a full application (CV, transcripts, and a short statement of interest) to with "[Open Positions]" in the subject line.