Xi Shen

Email: shenxiluc at gmail dot com

Bio: I am the Co-founder and Chief Scientist at Intellindust, where our team (Intellindust AI Lab) develops reliable and efficient computer vision algorithms for real-time applications on edge devices with limited NPU computation.

I completed my Ph.D. in the IMAGINE team at École des Ponts ParisTech. Before that, I was a Senior Researcher at Tencent AI Lab in Shenzhen, China.

If you’re interested in an internship focusing on safety AI or efficient training for real-time vision applications, feel free to reach out by email.

(Last update 09/11/2025)

GitHub CV Google Scholar

• 26 / 11 / 2025  SKEL-CF is now available on arXiv.
• 20 / 11 / 2025  SimROD: A Simple Baseline for Raw Object Detection with Global and Local Enhancements is accepted by AAAI 2026.
• 20 / 09 / 2025  Explicit visual prompting for universal foreground segmentations is accepted by TPAMI.
• 20 / 09 / 2025  "[DEIMv2] Real-Time Object Detection Meets DINOv3" is now available on arXiv.
• 01 / 07 / 2025  We took 2nd place in the NTIRE Joint Denoising and Demosaicing Challenge, 2nd place in the SoccerNet Ball Action Spotting Challenge, and received an honorable mention in the CVPR 2025 Foundation Few-shot Object Detection Challenge.
• 27 / 02 / 2025  "DEIM: DETR with Improved Matching for Fast Convergence" is accepted by CVPR 2025.
• 03 / 10 / 2024  "Designing nanotheranostics with machine learning" is accepted by Nature Nanotechnology 2024.
• 01 / 10 / 2024  Building on our CVPR work SURE, we won 1st place in the Open-set Recognition challenge at the Out Of Distribution Generalization in Computer Vision Workshop, ECCV 2024.
• 01 / 09 / 2024  We release COCO-FP, a dataset designed to evaluate false positive detections for COCO object detectors.
• 01 / 09 / 2024  "HTR-VT" is accepted by Pattern Recognition 2024.
• 01 / 06 / 2024  We achieved 1st place in Low-light Object Detection and 3rd place in Instance Segmentation at the 4th Physics-Based Vision meets Deep Learning Workshop, CVPR 2024.
• 27 / 02 / 2024  2/2 papers are accepted by CVPR 2024, thanks to all my co-authors!
• 01 / 08 / 2023  "TokenCut" is accepted by TPAMI.
• 01 / 07 / 2023  "LivelySpeaker" is accepted by ICCV 2023.
• 01 / 03 / 2023  3/3 papers are accepted by CVPR 2023, thanks to all my co-authors!
• 15 / 01 / 2023  Our paper T2M-GPT: Generating Human Motion from Textual Descriptions with Discrete Representations is available on arXiv.
• 22 / 11 / 2022  Our paper SadTalker: Learning Realistic 3D Motion Coefficients for Stylized Audio-Driven Single Image Talking Face Animation is available on arXiv.
• 30 / 08 / 2022  Our paper Back to MLP: A Simple Baseline for Human Motion Prediction (siMLPe) is accepted to WACV 2023.
• 14 / 07 / 2022  I am honoured to be acknowledged as Outstanding Reviewer for ICML 2022.
• 27 / 06 / 2022  Our paper Compressing Features for Learning with Noisy Labels is accepted to TNNLS 2022.
• 19 / 06 / 2022  CVPR 2022 workshop presentations: "TokenCut" in "L3D"; "SegSwap" in "Image Matching" and "Transformer".
• 02 / 03 / 2022  "TokenCut" is accepted to CVPR 2022. Check Gradio Demo here.
• 25 / 02 / 2022  Our paper "Self-Supervised Transformers for Unsupervised Object Discovery using Normalized Cut" (TokenCut) is published on arXiv.
• 13 / 01 / 2022  Our paper "Spatially-consistent Feature Matching and Learning for Heritage Image Analysis" is accepted to IJCV.
• 28 / 12 / 2021  I join Tencent AI Lab as a Senior Researcher, I will work with Dr. Jue Wang on Computer Vision.
• 03 / 12 / 2021  I am defending my Ph.D. today! [Slides] [Thesis]
• 29 / 10 / 2021  Our paper "Learning Co-segmentation by Segment Swapping for Retrieval and Discovery" is published on arXiv.
• 29 / 09 / 2021  Our paper "Re-ranking for image retrieval and transductive few-shot classification" is accepted to Neurips 2021.
• 18 / 04 / 2021  Our paper " Image Collation: Matching illustrations in manuscripts" is accepted to ICDAR 2021.
• 15 / 07 / 2020  Our paper " RANSAC-Flow: generic two-stage image alignment" is accepted to ECCV 2020. Watch the demo.
• 07 / 03 / 2020  The web application of historical watermark recognition is online!! Check the paper and the web application .
• 12 / 12 / 2019  Our paper " Empirical Bayes Transductive Meta-Learning with Synthetic Gradients" accepted to ICLR 2020.
• 06 / 08 / 2019  Our paper " Large-Scale Historical Watermark Recognition: dataset and a new consistency-based approach" is published on arXiv.
SHOW MORE...

• 🥇 1st Place — Open-set Recognition Challenge,
  Out-of-Distribution Generalization in Computer Vision Workshop, ECCV 2024
  (Based on our CVPR work SURE.)
• 🥇 1st Place — Low-light Object Detection,
  4th Physics-Based Vision Meets Deep Learning Workshop, CVPR 2024
• 🥉 3rd Place — Low-light Instance Segmentation,
  4th Physics-Based Vision Meets Deep Learning Workshop, CVPR 2024
• 🥈 2nd Place — Joint Denoising and Demosaicing,
  NTIRE Workshop, CVPR 2025
• 🥈 2nd Place — SoccerNet Ball Action Spotting,
  SoccerNet Challenges, CVPR 2025
• 🎖️ Honorable Mention — Foundation Few-shot Object Detection,
  Visual Perception via Learning in an Open World, CVPR 2025
• 🌟 Outstanding Reviewer, ICML 2022

^* indicates equal contribution.
⁺ indicates corresponding author.
Representative papers are highlighted.

Academic Services

        Reviewer for ICML, CVPR, ECCV, NeurIPS, ICLR, BMVC, WACV, TPAMI, PR, IJCV, ICCV.

Teacher Assistant

• 03 / 2021 - 04 / 2021   Université PSL (Paris Sciences & Lettres)  Digital Humanities Meet Artificial Intelligence (M1).
• 02 / 2021 - 06 / 2021   Ecole des Ponts ParisTech   Computer Vision for Mechanics of materials (L3).
• 02 / 2019 - 06 / 2019   Ecole des Ponts ParisTech   Traitement de l'information et vision artificielle (M1).

Sports

        I’m an amateur football player with a personal best of 11.30 seconds in the 100 meters. I also enjoy jogging, hiking, camping, and skiing.

Coding

        All of my released code is maintained on my GitHub account .