Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Human Identification at a Distance: Challenges, Methods and Results on the Competition HID 2025

Jingzhe Ma, Meng Zhang, Jianlong Yu, Kun Liu, Zunxiao Xu, Xue Cheng, Junjie Zhou, Yanfei Wang, Jiahang Li, Zepeng Wang, Kazuki Osamura, Rujie Liu, Narishige Abe, Jingjie Wang, Shunli Zhang, Haojun Xie, Jiajun Wu, Weiming Wu, Wenxiong Kang, Qingshuo Gao, Jiaming Xiong, Xianye Ben, Lei Chen, Lichen Song, Junjian Cui, Haijun Xiong, Junhao Lu, Bin Feng, Mengyuan Liu, Ji Zhou, Baoquan Zhao, Ke Xu, Yongzhen Huang, Liang Wang, Manuel J Marin-Jimenez, Md Atiqur Rahman Ahad, Shiqi Yu

TL;DR

HID 2025 advances gait-based human identification at a distance by evaluating on the challenging SUSTech-Competition dataset with new random splits to test cross-domain generalization. The top teams converge on strong backbones (notably DeepGaitV2) and emphasize data alignment, cross-dataset pretraining, pseudo-labeling, and re-ranking, with fine-tuning on the HID gallery proving particularly effective. The best result—94.2% rank-1 accuracy—establishes a new benchmark and highlights practical pathways for robust, domain-adaptive gait recognition. The study also identifies avenues for future work, including large pretrained models, multi-modal fusion, and privacy-aware synthetic data generation to push robustness in real-world deployments.

Abstract

Human identification at a distance (HID) is challenging because traditional biometric modalities such as face and fingerprints are often difficult to acquire in real-world scenarios. Gait recognition provides a practical alternative, as it can be captured reliably at a distance. To promote progress in gait recognition and provide a fair evaluation platform, the International Competition on Human Identification at a Distance (HID) has been organized annually since 2020. Since 2023, the competition has adopted the challenging SUSTech-Competition dataset, which features substantial variations in clothing, carried objects, and view angles. No dedicated training data are provided, requiring participants to train their models using external datasets. Each year, the competition applies a different random seed to generate distinct evaluation splits, which reduces the risk of overfitting and supports a fair assessment of cross-domain generalization. While HID 2023 and HID 2024 already used this dataset, HID 2025 explicitly examined whether algorithmic advances could surpass the accuracy limits observed previously. Despite the heightened difficulty, participants achieved further improvements, and the best-performing method reached 94.2% accuracy, setting a new benchmark on this dataset. We also analyze key technical trends and outline potential directions for future research in gait recognition.

Human Identification at a Distance: Challenges, Methods and Results on the Competition HID 2025

TL;DR

HID 2025 advances gait-based human identification at a distance by evaluating on the challenging SUSTech-Competition dataset with new random splits to test cross-domain generalization. The top teams converge on strong backbones (notably DeepGaitV2) and emphasize data alignment, cross-dataset pretraining, pseudo-labeling, and re-ranking, with fine-tuning on the HID gallery proving particularly effective. The best result—94.2% rank-1 accuracy—establishes a new benchmark and highlights practical pathways for robust, domain-adaptive gait recognition. The study also identifies avenues for future work, including large pretrained models, multi-modal fusion, and privacy-aware synthetic data generation to push robustness in real-world deployments.

Abstract

Human identification at a distance (HID) is challenging because traditional biometric modalities such as face and fingerprints are often difficult to acquire in real-world scenarios. Gait recognition provides a practical alternative, as it can be captured reliably at a distance. To promote progress in gait recognition and provide a fair evaluation platform, the International Competition on Human Identification at a Distance (HID) has been organized annually since 2020. Since 2023, the competition has adopted the challenging SUSTech-Competition dataset, which features substantial variations in clothing, carried objects, and view angles. No dedicated training data are provided, requiring participants to train their models using external datasets. Each year, the competition applies a different random seed to generate distinct evaluation splits, which reduces the risk of overfitting and supports a fair assessment of cross-domain generalization. While HID 2023 and HID 2024 already used this dataset, HID 2025 explicitly examined whether algorithmic advances could surpass the accuracy limits observed previously. Despite the heightened difficulty, participants achieved further improvements, and the best-performing method reached 94.2% accuracy, setting a new benchmark on this dataset. We also analyze key technical trends and outline potential directions for future research in gait recognition.
Paper Structure (18 sections, 8 figures, 2 tables)

This paper contains 18 sections, 8 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Dataset and Competition Details
  3. Dataset
  4. Performance metric
  5. Competition policies
  6. Competition statistics
  7. Methods of the Top Teams
  8. Team: BRAVO-FJ
  9. Team: BJTU-SSLL
  10. Team: SCUT-BIPLAB
  11. Team: :)
  12. Team: league
  13. Team: HUST-MCLAB
  14. Team: ITZY
  15. Team: sysu
  16. ...and 3 more sections

Figures (8)

  • Figure 1: Performance comparison of the top-10 final ranks across the last six HID competitions. The x-axis represents the final rank (e.g., 1st, 2nd). This visualization highlights the performance trends for each rank over the years. The orange-based bars show results for HID 2025 on the challenging SUSTech-Competition dataset, while the blue-based bars represent results on the CASIA-E dataset used from 2020-2022.
  • Figure 2: Some RGB images and their corresponding silhouettes from the dataset SUSTech-Competition. Many variations are included in the dataset.
  • Figure 3: The best scores and the numbers of submissions of each day during the competition.
  • Figure 4: The framework of Team BRAVO-FJ's method.
  • Figure 5: The framework of Team BJTU-SSLL's method.
  • ...and 3 more figures