search

AlibabaResearch / efficientteacher

A Supervised and Semi-Supervised Object Detection Library for YOLO Series
GNU General Public License v3.0
805 stars 147 forks source link

请问Single-class training中的NMS threshold during pseudo label generation如何调节呢？ #123

Open Wangxi404 opened 8 months ago

Wangxi404 commented 8 months ago

感谢优秀的工作和开源的代码🙏 目前希望把efficient teacher用来解决一个single-class detection的问题，看到在readme中提到的For example, if you are conducting single-class detection, then we suggest that the NMS threshold during pseudo label generation be reduced as much as possible。请问这个 NMS threshold during pseudo label generation是哪一个hyper parameters呢？如果不是hyper parameters，请问要怎么调节呢？ 谢谢🙏

Wangxi404 commented 8 months ago

除此之外，我还想请问为什么要更改这个NMS threshold呢？因为就算对于single class object detection，如果图片中有很多的instance，仍然需要大量的pseudo label。谢谢