simongraham / dsf-cnn

Dense Steerable Filter CNN

Dense Steerable Filter CNNs for Expoiting Rotational Symmetry in Histology Images

A densely connected rotation-equivariant CNN for histology image analysis.

Link to the pre-print.

NEWS: Our paper has now been published in IEEE Transactions on Medical Imaging. Find the published article here.

Getting Started

Environment instructions:

conda create --name dsf-cnn python=3.6
conda activate dsf-cnn
pip install -r requirements.txt

Repository Structure

• src/ contains executable files used to run the model. Further information on running the code can be found in the corresponding directory.
• loader/contains scripts for data loading and self implemented augmentation functions.
• misc/contains util scripts.
• model/class_pcam/ model architecture for dsf-cnn on PCam dataset
• model/seg_nuc/ model architecture for dsf-cnn on Kumar dataset
• model/seg_gland/ model architecture for dsf-cnn on CRAG dataset
• model/utils/ contains util scripts for the models.
• opt/ contains scripts that define the model hyperparameters and augmentation pipeline.
• config.py is the configuration file. Paths need to be changed accordingly.
• train.py and infer.py are the training and inference scripts respectively.
• process.py is the post processing script for obtaining the final instances for segmentation.

Citation

If any part of this code is used, please give appropriate citation to our paper.

@article{graham2020dense,
  title={Dense Steerable Filter CNNs for Exploiting Rotational Symmetry in Histology Images},
  author={Graham, Simon and Epstein, David and Rajpoot, Nasir},
  journal={arXiv preprint arXiv:2004.03037},
  year={2020}
}

Authors

See the list of contributors who participated in this project.

License

This project is licensed under the MIT License - see the LICENSE file for details

dsf-cnn's Issues

checkpoints for inference

Kindly provide with checkpoints to run inference file

Is this a bug?

Hello,
In the utils/rotation_utils.py line 236 and 237, the for loops are using the same ranges, is this right?

Best.

Data

where do i get the dataset you have mentioned

AssertionError: assert batch_size <= len(ds)

Hi , I got some error. Do you have any idea ? thanks

AssertionError: assert batch_size <= len(ds)

I am getting this error. I have created separate folders for training and validation data. Screenshots are attached below. Kindly help me solve this.

module metrics is not defined

module metrics is not defined in process.py

AssertionError

Dear Simon ,
I recently rectified the previous issues thanks for your feedback.
After solving the environment and hardware issues get this error. i will be glad to get your advice.

patch_extractor.py

Patch_extractor.py is a core code for extracting patches?The specific data set extraction patch should be implemented through the code written by oneself.

pytorch version release

Very thanks for your work. When will the pytorch version release?

How do you give the output shape to a 3 class classification

Typeerror

Hello Simon,
I get this error when i make changes you made.

Training problem

Hello sir,
I am trying to run your code in the training part but I have raised a confusing situation. It runs smoothly until it wants to start the training. When it want to start training epoch 1, the program just jump out but it didn't raise any error. Could you give me some inside thought about what happened here? I really appreciate it. Thank you.

Also, I have a question about the training data. When I download the MonuSeg dataset. It provides me .XML annotations and TIFF image. In you extract patches, you use PNG image and .NPY label. May you help me with how to convert the XML into the npy

?

Error

i get the following error, may i know how to fix it?

TypeError

Can anyone help with this error..?

tensorflow version Compatibility

I am using colab for experimentation. Here i cannot install tensorflow-gpu==1.12.0. Is there any way to the script with a newer version of TensorFlow? or Can you suggest me something regarding this issue?

PyTorch Version

Hey, very great work. But do you plan to release a PyTorch Version?

Best.