关于Neck的问题

[Lsz-20]

您好 我在尝试使用neck时发现报错：
KeyError：‘JPU’ is not in the {register}
KeyError：‘EncoderDecoder’ ：‘JPU’ is not in the models register
我检查了mmseg/models/init.py中的文件，似乎在_all_中缺少了关于neck的引入？
感谢回复！

[MengzhangLI]

Hi, could you use pip install -e . in your mmseg repo and try again?

The NECKS has been built here:

mmsegmentation/mmseg/models/builder.py

Line 25 in 2b2f107

return NECKS.build(cfg)

[Lsz-20]

Hi, could you use pip install -e . in your mmseg repo and try again?

The NECKS has been built here:

mmsegmentation/mmseg/models/builder.py

Line 25 in 2b2f107

return NECKS.build(cfg)

感谢您的回复，我查看了mmseg/model/下的init以及builder文件，确认和官方的并无差异，其中init为：

其中没有NECK的对吗？

[MengzhangLI]

Hi, could you use pip install -e . in your mmseg repo and try again?
The NECKS has been built here:

mmsegmentation/mmseg/models/builder.py

Line 25 in 2b2f107

return NECKS.build(cfg)

感谢您的回复，我查看了mmseg/model/下的init以及builder文件，确认和官方的并无差异，其中init为： 其中没有NECK的对吗？

Yep, actually I've checked my original repo re-implementing FastFCN, and it is the same as official code whose init does not have 'NECKS' in __all__.

[Lsz-20]

Yep, actually I've checked my original repo re-implementing FastFCN, and it is the same as official code whose init does not have 'NECKS' in __all__.

感谢您的回复~我重新安装了一下mmsegmentation，并检查了model/neck文件夹如下：

但似乎仍有以上错误？

[MengzhangLI]

Yep, actually I've checked my original repo re-implementing FastFCN, and it is the same as official code whose init does not have 'NECKS' in __all__.

感谢您的回复~我重新安装了一下mmsegmentation，并检查了model/neck文件夹如下： 但似乎仍有以上错误？

It is really wired. Can you write a detailed error report thus we could record it? Thanks!

[Lsz-20]

好的 感谢解答
报错如下：
Traceback (most recent call last):
File "/home/anaconda3/envs/py37_torch1.8/lib/python3.7/site-packages/mmcv/utils/registry.py", line 52, in build_from_cfg
return obj_cls(**args)
File "/home/lsz/Transformer/mmsegmentation/mmseg/models/segmentors/encoder_decoder.py", line 38, in init
self.neck = builder.build_neck(neck)
File "/home/lsz/Transformer/mmsegmentation/mmseg/models/builder.py", line 25, in build_neck
return NECKS.build(cfg)
File "/home/anaconda3/envs/py37_torch1.8/lib/python3.7/site-packages/mmcv/utils/registry.py", line 212, in build
return self.build_func(*args, **kwargs, registry=self)
File "/home/anaconda3/envs/py37_torch1.8/lib/python3.7/site-packages/mmcv/cnn/builder.py", line 27, in build_model_from_cfg
return build_from_cfg(cfg, registry, default_args)
File "/home/anaconda3/envs/py37_torch1.8/lib/python3.7/site-packages/mmcv/utils/registry.py", line 45, in build_from_cfg
f'{obj_type} is not in the {registry.name} registry')
KeyError: 'JPU is not in the models registry'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "tools/train.py", line 180, in
main()
File "tools/train.py", line 139, in main
test_cfg=cfg.get('test_cfg'))
File "/home/lsz/Transformer/mmsegmentation/mmseg/models/builder.py", line 49, in build_segmentor
cfg, default_args=dict(train_cfg=train_cfg, test_cfg=test_cfg))
File "/home/anaconda3/envs/py37_torch1.8/lib/python3.7/site-packages/mmcv/utils/registry.py", line 212, in build
return self.build_func(*args, **kwargs, registry=self)
File "/home/anaconda3/envs/py37_torch1.8/lib/python3.7/site-packages/mmcv/cnn/builder.py", line 27, in build_model_from_cfg
return build_from_cfg(cfg, registry, default_args)
File "/home/anaconda3/envs/py37_torch1.8/lib/python3.7/site-packages/mmcv/utils/registry.py", line 55, in build_from_cfg
raise type(e)(f'{obj_cls.name}: {e}')
KeyError: "EncoderDecoder: 'JPU is not in the models registry'"

感谢~

[Lsz-20]

It is really wired. Can you write a detailed error report thus we could record it? Thanks!
还有一个问题想要请教一下。看到您提供的neck是在fastfcn这个网络中使用的，请问其他网络（原先没有这个的）如果在configs文件中加入这个neck，是否可以正常使用呢？还是也会出现以上的‘JPU未被注册到模型中’的报错出现？
谢谢

[MengzhangLI]

Hi, theoretically it should work if input/output channels are aligned.

Could you try to use our provided FastFCN models here to testify whether JPU could be used in MMSegmentation?