search

PengyiZhang / SlimYOLOv3

This page is for the SlimYOLOv3: Narrower, Faster and Better for UAV Real-Time Applications
1.14k stars 320 forks source link

你好，可以提供vis2019转换完后的标签吗？ #27

Open 92ypli opened 5 years ago

92ypli commented 5 years ago

还有，怎么计算ap？

wydwww commented 5 years ago

You can use this script to convert visdrone dataset to darknet format and get the label.

Use test.py in https://github.com/ultralytics/yolov3 to test.

92ypli commented 5 years ago

非常感谢你的工作，不过我觉得你有一些实验细节做的并不好，你直接用darknet训coco的anchor，并没有重新做聚类，如果重新做聚类ap应该会提高很多。

wydwww commented 5 years ago

@wjjouc I didn't run the prune step.

wjjouc commented 5 years ago

@wydwww Thank you all the same. This problem has been bothering me for days.

KoapT commented 5 years ago

非常感谢你的工作，不过我觉得你有一些实验细节做的并不好，你直接用darknet训coco的anchor，并没有重新做聚类，如果重新做聚类ap应该会提高很多。

anchor size影响不大吧。

GinkgoX commented 4 years ago

How to calculate the AP results?