jiyuanfeng / mctrans Goto Github PK

你好，请问为什么有一类的指标一直是0，上不去。

我是用自己的数据集，使用模型mctrans_vgg32_d5.py，
显示：Traceback (most recent call last):
File "G:/codelearn/papercode329/3MCTrans-master/tools/train.py", line 128, in
main()
File "G:/codelearn/papercode329/3MCTrans-master/tools/train.py", line 103, in main
print(cfg.log_level)
File "g:\pythonproject\swin-zhenghe\mmcv-1.3.1\mmcv\utils\config.py", line 425, in getattr
return getattr(self._cfg_dict, name)
File "g:\pythonproject\swin-zhenghe\mmcv-1.3.1\mmcv\utils\config.py", line 44, in getattr
raise ex
AttributeError: 'ConfigDict' object has no attribute 'log_level'
请问有人知道什么问题吗，感谢

手动将log level设置为0

出现错误：
AttributeError: 'ConfigDict' object has no attribute 'data'

我怀疑是配置文件没弄好，有大哥知道嘛

I want to know more about the folding operation in the model, it is said that details can be found in the Appendix, but I cannot find the Appendix. Any idea? Thanks.

你好，请问为什么按照这个修改完，跑出来的最佳指标只有67.13

+---------------------------+--------+-------+-------+--------+-----------+--------+
| Class | Dice | Acc | IoU | Fscore | Precision | Recall |
+---------------------------+--------+-------+-------+--------+-----------+--------+
| Background | 142.16 | 600.0 | 245.8 | 142.16 | 80.63 | 600.0 |
| Neoplastic | 0.0 | 0.0 | 0.0 | nan | nan | 0.0 |
| Inflammatory | 0.0 | 0.0 | 0.0 | nan | nan | 0.0 |
| Connective | 0.0 | 0.0 | 0.0 | nan | nan | 0.0 |
| Dead | 0.0 | 0.0 | 0.0 | nan | nan | 0.0 |
| Non-Neoplastic Epithelial | 0.0 | 0.0 | 0.0 | nan | nan | 0.0 |
+---------------------------+--------+-------+-------+--------+-----------+--------+
这是我跑完模型它在log中统计的结果，为什么所有结果都是0？
而且我没有发现生成的预测图。

Is there a simple test pipeline to visualize the prediction, or i just need to follow the guidance in mmseg?
But it seems, with a custom model is cant be tested as in guidance.
for example as i rewrite the /envs/mctrans/lib/python3.7/site-packages/mmseg/models/builder.py, in your builder there is no build_segmentor method
Traceback (most recent call last):
File "/home/MCTrans/test/result_visualize_single.py", line 6, in
from mmseg.apis import init_segmentor, show_result_pyplot, inference_segmentor
File "/home/miniconda3/envs/mctrans/lib/python3.7/site-packages/mmseg/apis/init.py", line 2, in
from .inference import inference_segmentor, init_segmentor, show_result_pyplot
File "/home/miniconda3/envs/mctrans/lib/python3.7/site-packages/mmseg/apis/inference.py", line 9, in
from mmseg.models import build_segmentor
ImportError: cannot import name 'build_segmentor' from 'mmseg.models'

Is there a simple test pipeline?

Dear Author
I didn't see the use of simple flipping to enhance the training image in your code

Hi, thank you for your work. But I've noticed some inconsistencies with your paper:

1. It seems like the features extracted by the self-attention module are fed into the UNet decoder directly instead of the cross-attention module.
2. Proxy embedding is added to each input of the cross-attention layer, which seems not mentioned in the paper.

But maybe I misunderstood something? Look forward to your reply.

Hi!
Thanks for your code. I have a question about the Auxillary Loss. What's the gt for Auxillary Loss? Thank you very much!

首先感谢作者的工作！您提供的PanNuke数据集里面有split-images(masks)-npy和split-images(masks)，前者是.npy格式，后者是.png格式，请问使用这两种不同格式的数据集进行训练和测试是等价的吗？因为我load了npy格式的，发现其中的数据也是0~255的，和png格式的数据是一样的，是不是代表着使用.png格式的数据可以获得和.npy格式数据的同样效果？感谢！

Dear author：
什么时候公开多GPU训练的代码呢，想学习一下

How can I solve this problems？ Thanks you very much！！！

Traceback (most recent call last): File "Z:\py_item\7_MCTrans-master\MCTrans-master\tools\train.py", line 14, in <module> from mctrans.pipline import SegPipline File "Z:\py_item\7_MCTrans-master\MCTrans-master\mctrans\pipline\__init__.py", line 1, in <module> from .segpipline import SegPipline File "Z:\py_item\7_MCTrans-master\MCTrans-master\mctrans\pipline\segpipline.py", line 7, in <module> from mctrans.models import build_model File "Z:\py_item\7_MCTrans-master\MCTrans-master\mctrans\models\__init__.py", line 1, in <module> from .centers import * File "Z:\py_item\7_MCTrans-master\MCTrans-master\mctrans\models\centers\__init__.py", line 1, in <module> from .mctrans import MCTrans File "Z:\py_item\7_MCTrans-master\MCTrans-master\mctrans\models\centers\mctrans.py", line 8, in <module> from ..ops.modules import MSDeformAttn File "Z:\py_item\7_MCTrans-master\MCTrans-master\mctrans\models\ops\modules\__init__.py", line 9, in <module> from .ms_deform_attn import MSDeformAttn File "Z:\py_item\7_MCTrans-master\MCTrans-master\mctrans\models\ops\modules\ms_deform_attn.py", line 21, in <module> from ..functions import MSDeformAttnFunction File "Z:\py_item\7_MCTrans-master\MCTrans-master\mctrans\models\ops\functions\__init__.py", line 9, in <module> from .ms_deform_attn_func import MSDeformAttnFunction File "Z:\py_item\7_MCTrans-master\MCTrans-master\mctrans\models\ops\functions\ms_deform_attn_func.py", line 18, in <module> import MultiScaleDeformableAttention as MSDA ImportError: DLL load failed: 找不到指定的模块。

你好，我在运行程序时，出现了如下错误：
FileNotFoundError: file "/home/yyang17/projects/MCTrans/configs/debug/unet_vgg32_d5_256x256_400ep_pannuke.py" does not exist
在项目里面搜索 unet_vgg32_d5_256x256_400ep_pannuke.py 也没有发现，请问是否有部分模型文件没有上传呢？

Hello, i have a"Single channel image training error" problems。How can I solve that? Could I get your some suggestions？ Thank you very much！！！

The MCTrans link given in the readme section is not accessible

你好，我将这个模型用于别的数据集（19类），并更改了配置文件里的类别个数，但似乎并不能直接运行起来，出现了下面的错误：

Traceback (most recent call last):
  File "/mnt/csip-102/CaiJiaXin/MCTrans-master/train.py", line 131, in <module>
    main()
  File "/mnt/csip-102/CaiJiaXin/MCTrans-master/train.py", line 127, in main
    SegPipline(cfg, distributed, not args.no_validate).run()
  File "/mnt/csip-102/CaiJiaXin/MCTrans-master/mctrans/pipline/segpipline.py", line 99, in run
    self.runner.run(self.data_loaders, self.cfg.workflow, self.cfg.max_epochs)
  File "/opt/conda/lib/python3.7/site-packages/mmcv/runner/epoch_based_runner.py", line 127, in run
    epoch_runner(data_loaders[i], **kwargs)
  File "/opt/conda/lib/python3.7/site-packages/mmcv/runner/epoch_based_runner.py", line 50, in train
    self.run_iter(data_batch, train_mode=True, **kwargs)
  File "/opt/conda/lib/python3.7/site-packages/mmcv/runner/epoch_based_runner.py", line 30, in run_iter
    **kwargs)
  File "/opt/conda/lib/python3.7/site-packages/mmcv/parallel/data_parallel.py", line 75, in train_step
    return self.module.train_step(*inputs[0], **kwargs[0])
  File "/mnt/csip-102/CaiJiaXin/MCTrans-master/mctrans/models/segmentors/base.py", line 152, in train_step
    losses = self(**data_batch)
  File "/opt/conda/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
    return forward_call(*input, **kwargs)
  File "/opt/conda/lib/python3.7/site-packages/mmcv/runner/fp16_utils.py", line 109, in new_func
    return old_func(*args, **kwargs)
  File "/mnt/csip-102/CaiJiaXin/MCTrans-master/mctrans/models/segmentors/base.py", line 122, in forward
    return self.forward_train(img, img_metas, **kwargs)
  File "/mnt/csip-102/CaiJiaXin/MCTrans-master/mctrans/models/segmentors/encoder_decoder.py", line 92, in forward_train
    loss_aux = self._auxiliary_head_forward(x, seg_label, return_loss=True)
  File "/mnt/csip-102/CaiJiaXin/MCTrans-master/mctrans/models/segmentors/encoder_decoder.py", line 81, in _auxiliary_head_forward
    return self.aux_head.forward_train(x, seg_label)
  File "/mnt/csip-102/CaiJiaXin/MCTrans-master/mctrans/models/heads/mctrans_aux_head.py", line 49, in forward_train
    outputs = self.ca(inputs_flatten)
  File "/opt/conda/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
    return forward_call(*input, **kwargs)
  File "/mnt/csip-102/CaiJiaXin/MCTrans-master/mctrans/models/trans/transformer.py", line 145, in forward
    query = layer(query, src)
  File "/opt/conda/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
    return forward_call(*input, **kwargs)
  File "/mnt/csip-102/CaiJiaXin/MCTrans-master/mctrans/models/trans/transformer.py", line 122, in forward
    tgt2 = self.cross_attn(tgt, src, src)[0]
  File "/opt/conda/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
    return forward_call(*input, **kwargs)
  File "/opt/conda/lib/python3.7/site-packages/torch/nn/modules/activation.py", line 1038, in forward
    attn_mask=attn_mask)
  File "/opt/conda/lib/python3.7/site-packages/torch/nn/functional.py", line 5025, in multi_head_attention_forward
    k = k.contiguous().view(-1, bsz * num_heads, head_dim).transpose(0, 1)
RuntimeError: shape '[-1, 152, 16]' is invalid for input of size 2752512 

我看到关闭的问题中也有类似问题，但我按照他的修改方式也并不能成功，期待你的解答

hello
can the inference refer to https://github.com/JiYuanFeng/mmsegmentation/blob/master/demo/inference_demo.ipynb
thanks

When can the code be made public?

Hi,
Thanks for your great job, i like your algorithm very much.
i have been trying to reproduce your result, may i ask how big is your gpu memory, i have 16GB, but always "CUDA out of memory."

The implement of NonLocalUNet in this repo is VGG encoder + NonLocal center + UNet decoder, but the official implement is: replacing upsample and downsample with Global Aggregation Block, reduce the layers of UNet and replace the Concatenation with Add.
Refer to https://github.com/divelab/Non-local-U-Nets、https://arxiv.org/abs/1812.04103

Hello, I would like to know the meaning of the specific proxy embedding.

I use the pannuke dataset. But I found an error.
Traceback (most recent call last): File "tools/train.py", line 138, in <module> main() File "tools/train.py", line 135, in main SegPipline(cfg, distributed, not args.no_validate).run() File "/home/aMI_projects/MCTrans-master/mctrans/pipline/segpipline.py", line 99, in run self.runner.run(self.data_loaders, self.cfg.workflow, self.cfg.max_epochs) File "/home/anaconda3/envs/mctrans/lib/python3.7/site-packages/mmcv/runner/epoch_based_runner.py", line 127, in run epoch_runner(data_loaders[i], **kwargs) File "/home/anaconda3/envs/mctrans/lib/python3.7/site-packages/mmcv/runner/epoch_based_runner.py", line 50, in train self.run_iter(data_batch, train_mode=True, **kwargs) File "/home/anaconda3/envs/mctrans/lib/python3.7/site-packages/mmcv/runner/epoch_based_runner.py", line 30, in run_iter **kwargs) File "/home/anaconda3/envs/mctrans/lib/python3.7/site-packages/mmcv/parallel/data_parallel.py", line 75, in train_step return self.module.train_step(*inputs[0], **kwargs[0]) File "/home/aMI_projects/MCTrans-master/mctrans/models/segmentors/base.py", line 152, in train_step losses = self(**data_batch) File "/home/anaconda3/envs/mctrans/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl return forward_call(*input, **kwargs) File "/home/anaconda3/envs/mctrans/lib/python3.7/site-packages/mmcv/runner/fp16_utils.py", line 109, in new_func return old_func(*args, **kwargs) File "/home/aMI_projects/MCTrans-master/mctrans/models/segmentors/base.py", line 122, in forward return self.forward_train(img, img_metas, **kwargs) File "/home/aMI_projects/MCTrans-master/mctrans/models/segmentors/encoder_decoder.py", line 92, in forward_train loss_aux = self._auxiliary_head_forward(x, seg_label, return_loss=True) File "/home/aMI_projects/MCTrans-master/mctrans/models/segmentors/encoder_decoder.py", line 81, in _auxili ary_head_forward return self.aux_head.forward_train(x, seg_label) File "/home/aMI_projects/MCTrans-master/mctrans/models/heads/mctrans_aux_head.py", line 51, in forward_train outputs = self.ca(inputs_flatten) File "/home/anaconda3/envs/mctrans/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl return forward_call(*input, **kwargs) File "/home/aMI_projects/MCTrans-master/mctrans/models/trans/transformer.py", line 149, in forward query = layer(query, src) File "/home/anaconda3/envs/mctrans/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl return forward_call(*input, **kwargs) File "/home/aMI_projects/MCTrans-master/mctrans/models/trans/transformer.py", line 124, in forward tgt2 = self.cross_attn(tgt, src, src)[0] File "/home/anaconda3/envs/mctrans/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl return forward_call(*input, **kwargs) File "/home/anaconda3/envs/mctrans/lib/python3.7/site-packages/torch/nn/modules/activation.py", line 1045, in forward attn_mask=attn_mask, average_attn_weights=average_attn_weights) File "/home/anaconda3/envs/mctrans/lib/python3.7/site-packages/torch/nn/functional.py", line 5301, in multi_head_attention_forward k = k.contiguous().view(k.shape[0], bsz * num_heads, head_dim).transpose(0, 1) RuntimeError: shape '[32, 48, 16]' is invalid for input of size 22020096

Dear Author
I don't see the use of CA module in mctrans.py. Is the CA module useless?

bash make.sh
是不是需要cuda11？

Scource code cannot be aquired.