AttributeError: 'LayerNorm' object has no attribute 'affine'

Is the text annotation made by yourself? If I want to train my own dataset, do I need to make it？

Hi, congratulations on such a good work.
Could you please share the text annotation of MosMedData+ dataset?
Thanks.

Do the experimental results obtained in Table 2 of the paper use a pre-trained model? Using pre-training works well on the COVID-19 dataset, but the other one is not good.If not used, the opposite experiment is obtained.

How to segment multi-class objects?

I did not find the calculation code related to LV loss and EPI in the code, may I ask how to train under semi-supervised conditions?

您好！很抱歉再次打扰到您，我在复现您的代码的过程中总是被numpy版本冲突这个问题卡住，我尝试了先装bert,再装numpy还是不行，然后尝试了您之前所说的简单运行.train，我在运行过程中还是由于numpy版本冲突而停滞不前，我也尝试过创建一个新的虚拟环境，但仍然是在这个报错卡住，所以请求您能给你我一个具体的解决办法，谢谢您！

您好！很抱歉再次打扰到您，我想请问您已经提出3D模型了吗？我目前想做个3D的多模数据集，在您2D模型上面实现不了，另外我想请求您能告诉我一些公开的3D多模数据集，谢谢！

I've seen papers that say the MoNuSeg training set has 30 images and the test set contains 14 images? But the training data downloaded from your link has 37 images? May I ask which case do the experiments in this paper fit?

大佬好，我在复现论文的实验遇到了问题。我在win10平台下基于anaconda进行python环境配置。按照readme指引安装了numpy1.17.5。（numpy官网对1.17.5的python版本要求是3.5-3.8.）然后我在批量安装requirements.txt时出现了冲突，如下图所示

不同包对python版本的要求不兼容，请问大佬是如何处理这个问题的？

Hello, thank you for your all works.

I want to ask about that library from bert_embedding import BertEmbedding . this library returns an error. not loading the library. returns this error subprocess-exited-with-error.

How can i fix that error?

bert_embedding 只能兼容numpy1.14.6, 此版本与其它包不兼容，先安装bert_embedding 再装numpy也不行

你好！毋庸置疑您的论文写的非常出色，本人也相当想复刻你的代码，但是我在安装依赖时遇到了问题，按你所言python版本要为3.7，我在服务器上面的版本是3.8，我觉得这是兼容的，然后在直接安装,text时，报错numpy版本冲突，我尝试了先装bert-embedding再装.text，也试过先装bert-embedding，再装numpy==1.17.5，最后装.text，最后尝试了创建一个新的虚拟环境再来装这些依赖，但也同样报错，希望您能教教我，该怎么一步一步正确安装依赖，谢谢您！

Hi,
The link you provided (http://medicalsegmentation.com/covid19) contains 100 axial CT images+9 volumes+20 volumes, but MosMedData contains 50 annotated volumes. I did not figure out why a smaller dataset (MosMedData+) includes a larger dataset (MosMedData). Can you provide any reference that describes MosMedData+ dataset and claims that it includes MosMedData?

Thanks.

Dear Zihan,

Thank you for your amazing work. I am currently attempting to replicate your study for further research. However, I've encountered an issue with the dataset formats for QaTa-COV19 and MosMedData+.

To address these concerns, I have sent a detailed email outlining the specific issues encountered with the dataset formats.

Your feedback would be greatly appreciated, and I eagerly await your response. Thank you very much for your time and consideration.

Best regards,
Pengyu Zhao

Hello!! Thank you very much for your work, LViT paper is very helpful to me!!

Here I would like to ask about my results of replicating QaTa-COV19 data set on LViT model.

I used the QaTa-COV19 data set provided by you. The training set is 5716 pieces and the verification set is 1429 pieces. The division of the training set and the verification set according to the text.xlsx file you provided.

The Batch_size is set to 24, the learning rate is set to 3e-4, the learning rate is reduced to 1e-4 minimum after 50 epochs, if the effect does not improve, then stop training

I didn't make any changes to the original code I downloaded, and the random torrent uses 666 from the original code

However, the Dice value I reproduced was 81.97, which was slightly different from the result recorded in the paper

The device I am using is RTX 3090 24GB video memory and the Python version is 3.7.2
I have sent you an email and I am looking forward to your reply！！

I congratulate you for your work.
I would like to ask about your work, can I access the pth.tar checkpoint to work pre-train with the MonuSeg dataset? And how can I run this model without the text data? Thanks

1,MoNuSeg数据集处理和UCTransnet是否一致？（看了下数据结构，应该差不多一致）
2.关于文章的对比试验表格数据，作者您使用了UCTransnet的best_model去测试的结果dice：0.7987？那么其他模型的您从哪里得到？比如，TransUnet,Unet,Unet++。
3，Uctransnet作者回复由于随机性，训练的结果会有偏差，然后给出自己的最好的best_model,我向其要了一份，测试结果和您的0.7987大差不离；然而我是用Unet---也就是UCTransnet作者写的代码，随机seed训练，最好的结果有78%以上的，为什么您的只有76%？而且他的也不高。

Your work is very remarkable but while trying to run the code on my local machine
I am unable to run it
I have created a virtual conda environment and while trying to run conda install --file requirements.txt it shows many packages are not available in the current channels.So how should I run the code ?? while normal pip installing for bert-embedding It shows a legacy error .So if you could help me out

Hello.
Using language to segment medical Image was so interesting.
I want to train the model for myself.

So when can I get the text annotation QaTa-COV19 Dataset?
Can I receive the model early?

Thanks.

Hi, I am very impressed by your work!

I would like to know when your pre-trained model will be updated.

Thank you

Hello, and thank you so much for your fantastic work. I have a question about the test_model.py file. Regardless of whether I pasted the entire path to my c directory with the dataset name, I always get this issue when I try to run this program. I am unable to fix the problem, though. Please provide me with some advice, and I appreciate it in advance. I may be experiencing problems with my packages, or possibly the issue is actually with the path directory.

C:\Users\lenovo\anaconda3\envs\trans\python.exe C:\Users\lenovo\PycharmProjects\Lvt\test_model.py
Traceback (most recent call last):
File "C:\Users\lenovo\PycharmProjects\Lvt\test_model.py", line 78, in
checkpoint = torch.load(model_path, map_location='cuda')
File "C:\Users\lenovo\anaconda3\envs\trans\lib\site-packages\torch\serialization.py", line 581, in load
with _open_file_like(f, 'rb') as opened_file:
File "C:\Users\lenovo\anaconda3\envs\trans\lib\site-packages\torch\serialization.py", line 230, in _open_file_like
return _open_file(name_or_buffer, mode)
File "C:\Users\lenovo\anaconda3\envs\trans\lib\site-packages\torch\serialization.py", line 211, in init
super(_open_file, self).init(open(name, mode))
FileNotFoundError: [Errno 2] No such file or directory: 'C:/Users/lenovo/PycharmProjects/Lvt/MoNuSeg/LViT/test_session_05.24_16h30/models/best_model-LViT.pth.tar'

requirements.txt中没有bert-embedding，且bert-embedding只与numpy 1.14.6兼容，会产生冲突，该如何解决

Hello. Thank you very much for sharing the text annotation of MosMedData+Dataset. But I am confused about how to match the images of the dataset with text annotations.
Specifically, there are several types of image naming in your text annotation: 'bjorke'_ X ',' Jun_ Coronacases_ CaseX ',' Jun_ Radiopaedia_ X ',' Morozov_ Study_ X '.
From my guess, 'bjorke'_ X ' may corresponds to' COVID-19 CT segmentation dataset ', while' Jun '_ Coronacases_ CaseX 'and' Jun_ Radiopaedia_ X ' may corresponds to' COVID-19-CT-Seg_ 20 cases'. then which image data does 'morozov study ' correspond to?
I'm sorry for being confused about this. Can you provide detailed instructions on how to match them?

Hello, your paper is interesting.
I have some questions about Exponential Pseudo-label Iteration mechanism, but I didn't find the code for this part. Did you push it on your repo?

Can this setup be taken into consideration for UNet series models? If not, how can a new UNet series configuration be created? or is this only applicable to transformers?

##########################################################################

CTrans configs

##########################################################################
def get_CTranS_config():
config = ml_collections.ConfigDict()
config.transformer = ml_collections.ConfigDict()
config.KV_size = 960 # KV_size = Q1 + Q2 + Q3 + Q4
config.transformer.num_heads = 4
config.transformer.num_layers = 4
config.expand_ratio = 4 # MLP channel dimension expand ratio
config.transformer.embeddings_dropout_rate = 0.1
config.transformer.attention_dropout_rate = 0.1
config.transformer.dropout_rate = 0
config.patch_sizes = [16, 8, 4, 2]
config.base_channel = 24 # base channel of U-Net
config.n_classes = 1
return config

Hello, could you tell me the proportion of covid-19 training set, verification set and test set? Is it random?

Dear zihan:

I hope this message finds you well. Firstly, I want to extend my gratitude for your innovative work on LViT. It has been a valuable resource, and I have successfully obtained results using the provided methods.

However, I am currently encountering a challenge with training LViT models without using text. I referred to your previous responses and attempted a modification based on this screenshot from your GitHub repository:

I modified the code as follows, as per my understanding:

Unfortunately, this change did not yield the expected results, and I encountered the following error:

Could you please provide some guidance or additional details on the correct approach to train LViT models without text? Any advice or further clarification you can offer would be greatly appreciated.

Thank you for your time and consideration. I look forward to your valuable input.

Best regards,
Pengyu

Hi, congratulations on such a good work.
Could please share the pre-trained models for MonuSeg dataset.

Can you release the EPI code part now? This is a core part of the paper, we appreciate you can release it so we can reproduce the result. Thanks.

Hello, What kind of optimisation can I do to train on 1 gpu?

Hello, where is the text description of the test set of QaTa-COV19?

Hi,

The bert-embedding and gluonnlp packages are depreicated, therefore cannot be installed. Are there other ways to achieve same result reported in the paper?

大佬能分享下数据集供我学习学习吗，感激不尽，我的邮箱 [email protected]

李老师你好，我是重庆邮电大学的研究生，看到了你这篇LViT论文，写得很好，但是我有些不懂的地方，所以想询问你一下几个问题？就是这篇论文中数据集，验证集和训练集的划分标注。covid19对比标签是什么，文中不同结构形式的文本构成对比标签，这个是怎么构成的呢？还请您解答一下，谢谢老师！！

Hi,
I have a question about the MosMedData+ dataset in your LViT paper. The link you provided(http://medicalsegmentation.com/covid19) is not MosMedData, while the correct link should be (https://academictorrents.com/details/f2175c4676e041ea65568bb70c2bcd15c7325fd2/tech&filelist=1). Did you use the wrong name or wrong link?
Thanks!

Dear zihan,

Thank you for your impressive work!

I have been following your work and I have successfully run the code with the setting 'model_type="LViT"', which does not require a pretrained model. However, I am facing some challenges in understanding how to obtain the pretained model for the setting 'model_type="LViT"'.

As discussed in Section 2.1 of your instructions:

It appears that a U-Net model might be a prerequistite for the LViT model. Could you please clarifiy if this is the case? If so, does this imply I need to train a U-Net model first before proceeding with 'LViT_pretrain'? Also, in such a scenario, should I change the model type and write the corresponding code here?

Furthermore, I am also curious about how to load the pretrained U-Net model once it is obtained. Is this U-Net model directly applicable to LViT_pretrain, or are there additional steps or modifications required?

Your guidance on these matters would be greatly appreciated, as it would greatly assist me in understanding and utilizing your work more effectively.

Thank you for your time and consideration. I am looking forward to your valuable insights.

Best regards,
Pengyu Zhao

Looking forward to your reply
Thanks！

Dear Zihan,

I hope this message finds you well. As I delve deeper into the LViT model, I have encountered a specific aspect of the network structure that I am eager to understand better with your guidance.

In my thorough examination of the LViT's network architecture, I observed that the PixLevelModule, as defined in the code, appears not to be used in the forward pass. To better illustrate my point, I am referencing the following sections of the code from your GitHub repository:

This observation has led to some confusion about its role in the overall network structure as described in your paper. Could you kindly provide clarification on this and, if possible, share the correct code that reflects the implementation used in the paper?

Your insights are invaluable to my understanding and proper implementation of the LViT model. I greatly appreciate your time and the support you've provided thus far.

Thank you once again for your assistance.

Best regards,
Pengyu

Hi there,

I am trying to reproduce your results on QaTa-COV19 Dataset. But I am wondering how you do the Train-Val split on QaTa-COV19 Dataset. Could you release more details about that? Thanks.

Also looking forward to the text annotations on this dataset.