search

datvuthanh / HybridNets

HybridNets: End-to-End Perception Network
MIT License
594 stars 121 forks source link

Evaluate the model ? #35

Open huoguangdiandian opened 2 years ago

huoguangdiandian commented 2 years ago

hi, I have two problems and hope to help me out, thank you! 1) I have trained new model and want to evaluate the model. There are two functions val() and val_from_cmd(), in val.py. what's the difference between the two functions? 2) I have modified the number of categories to 10, and when I run the val.py, there is a error:ValueError: operands could not be broadcast together with shapes (12,) (4,) image

xoiga123 commented 2 years ago
  1. val_from_cmd is a cut-down version of val, to validate from command-line instead of from training. I think in this stage of the repo, trading some repetitive code for better readability is okay. Maybe we'll merge them with some if-else in the future.

  2. 20, we haven't supported segmentation other than road and lane line yet. I have some free time now so I'll work on it.

huoguangdiandian commented 2 years ago

@xoiga123 your means that I can not use val.py to evaluate the model yet?

xoiga123 commented 2 years ago

@huoguangdiandian yes, yet :running_man:

huoguangdiandian commented 2 years ago

Thank you for your reply

xoiga123 commented 2 years ago

@huoguangdiandian False alarm, it was not related to segmentation classes at all. It was about batch-size thingy. Pull the repo and try again, I also merged val_from_cmd with val.

huoguangdiandian commented 2 years ago

@xoiga123 I only use the test results, partition is not used, val_from_cmd is available. I have another problem is that Is there a pre-training model based on training 10 classes???

xoiga123 commented 2 years ago

@huoguangdiandian Sorry, I don't quite understand your first sentence. I fixed your bug in the latest commit. Save your 10-class weight and clone the repository again and validate on that, it should work then. And no, we currently don't have a pre-trained model on 10 classes.