ArticleMetadataMain page

source · 2026-05-23

SoccerNet-Tracking足球多目标跟踪数据集与基准

Venue / source
CVPR Workshop 2022
Published
2022-04-14
Ingested
2026-05-23
Source type
paper
Paper pageopenaccess.thecvf.com · Cioppa_SoccerNet-Tracking_Multi…PDFopenaccess.thecvf.com · Cioppa_SoccerNet-Tracking_Multi…Code repositorygithub.com · sn-tracking
#near-cvpr-2025#video-understanding#tracking#benchmarking#primary-source

一句话结论

SoccerNet-Tracking 给足球视频里的球员、裁判和球建立了可复用的 MOT 数据集与 benchmark，是 topics/sports-ai-roadmap 从 SoccerNet 事件生态走向 full-pitch game-state reconstruction 的底层 tracking 证据。

论文定位

这篇论文提出 SoccerNet-Tracking：200 段 30 秒挑战片段，加上一段完整 45 分钟半场长时跟踪视频；标注包含 bounding boxes 和 tracklet IDs，目标覆盖 players、referees 和 ball。它解决的是足球场景里最基础但很难绕开的对象连续性问题：快速运动、遮挡、相似球衣、球的小目标属性和长时 re-identification。

对体育 AI 子线的价值

来源可靠性与可溯源性

  • 来源层级：CVPR Workshop 2022 / CVSports primary paper；CVF OpenAccess 页面、CVF PDF、arXiv 页面和官方 SoccerNet tracking repo 均可访问。
  • 可溯源材料：本地已保存 abstract.md、paper.pdf、paper-text.md、meta.yaml 与 links.yaml；PDF 本地从 arXiv mirror 保存，SHA256：f3cb64b8a707960afa4db8d53db3bd00c4ef119dd75dbc2e53a22fffc1c70806。
  • 使用边界：适合作为 football MOT / SoccerNet-GSR 前置数据层的强 primary evidence；它本身还停在 bbox-level tracking，pitch coordinates、身份语义和战术解释需要后续 GSR / trajectory sources 承接。

和现有证据的关系

它补在 sources/2026-05-05-soccernet-2023-tracking-mot4mot 和 sources/2026-04-25-soccernet-game-state-reconstruction 之间：前者是 challenge 系统样本，后者把 tracking 扩展到 minimap / pitch localization / role-team-jersey identity。SoccerNet-Tracking 则提供最原始的足球 MOT benchmark substrate。

局限或疑问

  • 论文强调 fast motion 与 severe occlusion 下仍有明显提升空间，这正好解释了为什么 sources/2026-05-23-broadcast2pitch-game-state-reconstruction 这类后续工作需要更强的几何对齐、ReID 和 tracklet refinement。
  • 数据集仍以足球转播片段为主，跨联赛、跨画质、跨机位泛化需要继续用 TeamTrack / SportsMOT / GSR follow-up 验证。

原始材料

  • raw/ingest/2026-05-23-soccernet-tracking/paper.pdf
  • raw/ingest/2026-05-23-soccernet-tracking/paper-text.md
  • raw/ingest/2026-05-23-soccernet-tracking/abstract.md
  • raw/ingest/2026-05-23-soccernet-tracking/links.yaml

相关页面

Metadata

{
  "id": "2026-05-23-soccernet-tracking",
  "type": "source",
  "title": "SoccerNet-Tracking：足球多目标跟踪数据集与基准",
  "status": "reviewed",
  "created": "2026-05-23",
  "updated": "2026-05-23",
  "venue": "CVPR Workshop 2022",
  "published_at": "2022-04-14",
  "ingested_at": "2026-05-23",
  "tags": [
    "near-cvpr-2025",
    "video-understanding",
    "tracking",
    "benchmarking",
    "primary-source"
  ],
  "note_status": "reviewed",
  "source_type": "paper",
  "authors": [
    "Anthony Cioppa",
    "Silvio Giancola",
    "Adrien Deliège",
    "Le Kang",
    "Xin Zhou",
    "Zhiyu Cheng",
    "Bernard Ghanem",
    "Marc Van Droogenbroeck"
  ],
  "canonical_links": [
    "https://openaccess.thecvf.com/content/CVPR2022W/CVSports/html/Cioppa_SoccerNet-Tracking_Multiple_Object_Tracking_Dataset_and_Benchmark_in_Soccer_Videos_CVPRW_2022_paper.html",
    "https://openaccess.thecvf.com/content/CVPR2022W/CVSports/papers/Cioppa_SoccerNet-Tracking_Multiple_Object_Tracking_Dataset_and_Benchmark_in_Soccer_Videos_CVPRW_2022_paper.pdf",
    "https://arxiv.org/abs/2204.06918",
    "https://github.com/SoccerNet/sn-tracking"
  ],
  "raw_entry": "raw/ingest/2026-05-23-soccernet-tracking/",
  "topics": [
    "topics/sports-ai-video-understanding",
    "topics/sports-ai-roadmap",
    "topics/video-understanding"
  ],
  "entities": [
    "entities/sportsmot"
  ],
  "claims": [],
  "questions": []
}