search

qfgaohao / pytorch-ssd

MobileNetV1, MobileNetV2, VGG based SSD/SSD-lite implementation in Pytorch 1.0 / Pytorch 0.4. Out-of-box support for retraining on Open Images dataset. ONNX and Caffe2 support. Experiment Ideas like CoordConv.
https://medium.com/@smallfishbigsea/understand-ssd-and-implement-your-own-caa3232cd6ad
MIT License
1.39k stars 529 forks source link

[Question] what these parameters mean? #174

Closed unanchoi closed 2 years ago

unanchoi commented 2 years ago

Hi. I have question about input parameters

  1. What is difference between --base_net and --pretrained_ssd
  2. What is difference between --freeze_base_net and --freeze_ner
  3. what does mean the parameter 'debug_step' in training

https://github.com/qfgaohao/pytorch-ssd/blob/f61ab424d09bf3d4bb3925693579ac0a92541b0d/train_ssd.py#L41-L67

SarBH commented 2 years ago

This is a guess: --base_net its the ability to provide a pretrained base (the classifier backbone part (vgg16, mobilenetV1, etc) --pretrained_ssd providing a pretrained complete SSD model (base and SSD portions) If you look in the readme there are some urls of models, some of them have "ssd", and i guess you should pass them with --pretrained_ssd others dont have "ssd" so you would need to pass via --base_net And depending on wether u want to train the SSD portions (except the head because that's num_classes specific) you choose what to freeze

unanchoi commented 2 years ago

thank you for comment!