search

Visual-Attention-Network / SegNeXt

Official Pytorch implementations for "SegNeXt: Rethinking Convolutional Attention Design for Semantic Segmentation" (NeurIPS 2022)
Apache License 2.0
794 stars 85 forks source link

Cannot reproduce your results #36

Closed il1um closed 2 years ago

il1um commented 2 years ago

hi，i'm sorry to bother you but when i use your local_config(default setting) to train neither using cityscape nor ade20k can reproduce your results. Do you have any idea about this?

When i directly use the pretrained .pth file that you provide，i can get the result 79.25%miou on cityscapes. But when I train the model locally, the result is terrible.

Here is my result :60.7% miou on cityscapes; 18.16%miou on ADE20K

I would be very grateful if you could reply me ASAP.

uyzhang commented 2 years ago

Check whether the pre-training weight is loaded into the network.

il1um commented 2 years ago

Thank you so much for your quick reply，but in my experiment, with pre-training weight and without pre-training weight can both only get around 18% miou on ADE20K. (w/o:18.48%; w:18.16%)

here are the questions: 1: how can i check if i really use or don't use the pre-training weight? 2：Is the score of 18.48%miou on ADE20K normal (correct) if I don't use pre-training weights?

Thanks

uyzhang commented 2 years ago
  1. You can check the log to see whether to load it.
  2. I did not try not to load the mIoU of the training pre-training model.