problem of the weight does not match the network structure

[WangGeStu]

maybe you can not match the network structure and weight.
because the original weight model has been deleted.
you can get new weight model from http://sceneparsing.csail.mit.edu/model/pytorch/ade20k-resnet50dilated-ppm_deepsup/
but this weight can not match with the code in semantic_features .
you should update and change the code from https://github.com/CSAILVision/semantic-segmentation-pytorch/tree/master/mit_semseg

by change the code ,you can match the network structure and weight ,then you can run.

[xinli2008]

hello,The link has expired! can you provide a new link?
https://github.com/CSAILVision/semantic-segmentation-pytorch/tree/master/mit_semseg

[WangGeStu]

there is no failure

[xinli2008]

好的，可能是刚才我的网络出了问题，一直连不上去。对了，请问你这个代码最后跑出结果了吗？因为我看之前人的回复，他们说自己跑出的结果和论文给出的结果很不一样，我也遇到过这个问题。

[WangGeStu]

现在已经找不到他们之前用的模型权重了，所以实验结果我没有对比过。我只是把它作为一个工具在用，还不知道效果如何。