跳到正文
返回论文列表
cs.CV提交于 已译

行人跟踪中的持久身份:在线注册与注册识别

Persistent Identity in Person Tracking: Online Enrollment and Registered Recognition

Ratha Siv · Matei Mancas · Bernard Gosselin

中文摘要

将 PoseTReID 扩展为带在线外观库匹配(Gallery)和残差冲突防护的注册 SVM 识别,所有方法均使用给定的真实(GT)行人框。全框评估的 IDF1/HOTA 在各视频间等权平均。在六个 SportsMOT 片段上,StrongSORT 领先独立跟踪器(79.44/84.72%),并超过两个 Gallery 变体。在十一个视频上,Gallery AIN(83.29/85.44%)和 COPE(84.66/86.38%)在结果完整时超过独立跟踪器;该领先依赖两个事后加入的 GTA 开发集,在原始九个视频上不存在。注册 SVM 在完整结果中领先两组,AIN 在十一个视频上达 95.72/94.82%。它们使用约 20% 帧的全部 GT 裁剪,包括未来出现。在 MOT20-03 上,注册 AIN 达到 96.80% IDF1,并为 90.72% 的评估 GT 观测分配正确注册名。Gallery 和独立跟踪器没有逐视频的带标签注册。两种身份路线都能跨内部标题卡恢复身份,但 Gallery 在三个视频上落后于 SORT(HOTA 上为四个),两个 SVM 在 MOT16-04/MOT20-03 上落后于 SORT。在六片段仅持久评分下,原生 Gallery 比匹配对照平均 IDF1 高 2.39 个百分点,但正确 P 覆盖率更低,编码器尝试次数为 5.02 倍。姿态门控效果混合,防护的指标效应未被单独分离。

关键要点

  1. 01扩展 PoseTReID,加入在线外观库匹配与注册 SVM 识别的残差冲突防护。
  2. 02Gallery 在十一个视频上超过独立跟踪器,但领先依赖事后加入的 GTA 开发集。
  3. 03注册 SVM 在完整结果中领先,AIN 在 MOT20-03 达 96.80% IDF1。
  4. 04Gallery 在部分视频上落后 SORT,且编码器尝试次数为对照的 5.02 倍。
  5. 05姿态门控效果混合,防护的指标效应未被单独分离。

解读

尚无解读。

原始英文摘要

arXiv:2611.25283v1 Announce Type: new Abstract: We extend PoseTReID with online appearance-bank matching (Gallery) and a residual-conflict guard for registered SVM recognition. All methods use supplied ground-truth (GT) person boxes. We average all-box assessment IDF1/HOTA equally across videos. On six SportsMOT clips, StrongSORT leads standalone trackers (79.44/84.72%) and exceeds both Gallery variants. Over eleven videos, Gallery AIN (83.29/85.44%) and COPE (84.66/86.38%) exceed standalone trackers with complete results. This lead depends on two retrospectively added GTA development compilations and is absent on the original nine videos. Registered SVMs lead both cohorts among complete results, with AIN at 95.72/94.82% over eleven videos. They use every GT crop from approximately 20% of frames, including future appearances. On MOT20-03, registered AIN reached 96.80% IDF1 and assigned the correct enrolled name to 90.72% of assessment GT observations. Gallery and standalone trackers receive no per-video labelled enrollment. Both identity routes restore identities across internal title cards, but Gallery trails SORT on three videos (four in HOTA) and both SVMs trail SORT on MOT16-04/MOT20-03. Under six-clip persistent-only scoring, native Gallery gains 2.39 percentage points in mean IDF1 over a matched control, with lower correct-P coverage and 5.02 times as many encoder attempts. Pose gating is mixed; the guard's metric effect is not isolated.

同方向论文 · cs.CV

查看全部 →