The PR Curve when the loop distance less than 3m.

[neng-wang]

Dear author @ChongjianYUAN ,

Thank you for sharing your outstanding work. STD is user-friendly and performs well from visualization in rviz.

However, when I test the STD and consider a loop pair as true positives if their Euclidean distance is less than 3m, differing from the paper's criteria (20m for true positives). I find STD's performance degrades significantly (as shown in the figure).

We suspect parameter tuning is key, but despite adjustments, STD does not yield satisfactory results.

Could you advise on parameter tuning or other methods to enhance STD's performance with a 3m true positive range?

[stupidme123]

你好，你是否测试了的20m的数据呢？我即便使用20m真值阈值也无法得到论文中的曲线