search

WongKinYiu / yolov7

Implementation of paper - YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors
GNU General Public License v3.0
13.36k stars 4.22k forks source link

What is the difference between yolov7.yaml in cfg/deploy and cfg/training #74

Open lixiangyussss opened 2 years ago

lixiangyussss commented 2 years ago

What is the difference between yolov7.yaml in cfg/deploy and cfg/training

YaoQ commented 2 years ago

between Detection and IDetection function part

lixiangyussss commented 2 years ago

between Detection and IDetection function part Which one has higher accuracy, how much higher, do I need to replace v7.yaml in cfg/training and retrain to represent yolov7 with new results?

WongKinYiu commented 2 years ago

Models in cfg/training are training-time models, after finish training, we re-parameterized them into inference-time models in cfg/deploy.

lixiangyussss commented 2 years ago

cfg/training 中的模型是训练时模型，训练完成后，我们将它们重新参数化为 cfg/deploy 中的推理时模型。 Ok thanks, then if I want to compare the experimental results with v7 in my paper, what should I do?

WongKinYiu commented 2 years ago

Both have same results, if you want to reproduce training procedure, please use cfg/training.

lixiangyussss commented 2 years ago

两者的结果相同，如果要重现训练过程，请使用 cfg/training。

Got it. All right. Thank you

mickey-cool commented 2 years ago

来自台湾的朋友，Would you please tell us how to use cfg/deploy,means how to re-parameterize? Thank you!