marrlab/instantdl

InstantDL: An easy and convenient deep learning pipeline for image segmentation and classification

InstandDL: An easy and convenient deep learning pipeline for image segmentation and classification

InstantDL enables experts and non-experts to use state-of-the art deep learning methods on biomedical image data. InstantDL offers the four most common tasks in medical image processing: Semantic segmentation, instance segmentation, pixel-wise regression and classification. For more in depth discussion on the methods, as well as comparing the results and bechmarks using this package, please refer to our preprint on bioRxiv here

---

Documentation

For documentation please refere to docs

For a short video introducing InstantDL please see:

Contributing

We are happy about any contributions. For any suggested changes, please send a pull request to the develop branch.

Citation

If you use InstantDL, please cite this paper:

@article {
author = {Waibel, Dominik Jens Elias and Shetab Boushehri, Sayedali and Marr, Carsten},
title = {InstantDL - An easy-to-use deep learning pipeline for image segmentation and classification},
year = {2021},
doi = {10.1186/s12859-021-04037-3},
URL = {https://bmcbioinformatics.biomedcentral.com/articles/10.1186/s12859-021-04037-3#article-info},
eprint = {https://doi.org/10.1186/s12859-021-04037-3},
journal = {BMC Bioinformatics}
}

instantdl's Issues

_config.yml seems to be empty

_config.yml should contain packages to be installed, right?

google colab example is not complete

the google colab example does not look correct. Needs probably to be checked

Iterations = 0 with weights in another directory

When running InstantDL with Iterations_over_dataset = 0 and the weights not saved in the "/data directory/logs/" directory it will give crash as it searches for the weights in the wrong folder.

GetPipeline error

When I run InstantDL with:

from instantdl import GetPipeLine

configs = {"use_algorithm": "SemanticSegmentation","path": "/home/dominik/Documents/InstantDL_dev_lsd/Vessel12_small2/","batchsize": 1,"iterations_over_dataset": 10,"data_gen_args": {},"loss_function": "mse"}

pipeline = GetPipeLine(configs)

I get the following error:
TypeError: GetPipeLine() missing 10 required positional arguments: 'path', 'pretrained_weights', 'batchsize', 'iterations_over_dataset', 'data_gen_args', 'loss_function', 'num_classes', 'image_size', 'calculate_uncertainty', and 'evaluation'

Running it in PyCharm is no problem at all

Bug in image resize

In SemanticSegmentation.py
Line 46 should be:
if self.image_size == False:

Missing data organization figure

In the README file at: https://github.com/marrlab/InstantDL/tree/develop/instantdl#how-to-use-the-code, the following sentence:

"The pipeline is based on folders. Please put your data manually in the corresponding folders as illustrated by the figure above. The folder names must not be changed as this will stop the pipeline from functioning.", says that the data organization can be viewed using a figure. The figure is not found in the README.

pattern design and documentation

We should follow a better pattern design and documentation. My suggestion would be to follow for example skimage as an example

CPU multiprocessing on Windows

In the code comments (here) the issue of multiprocessing on windows is already described. However, the workers are never passed to keras_model.fit_generator (see here).

Uncommenting the line 2385 in the RCNNmodel solved the issue for me.

Produced Errors:
Exception in thread Thread-9: Traceback (most recent call last): File "C:\ProgramData\Anaconda3\lib\threading.py", line 926, in _bootstrap_inner self.run() File "C:\ProgramData\Anaconda3\lib\threading.py", line 870, in run self._target(*self._args, **self._kwargs) File "C:\ProgramData\Anaconda3\lib\site-packages\keras\utils\data_utils.py", line 666, in _run with closing(self.executor_fn(_SHARED_SEQUENCES)) as executor: File "C:\ProgramData\Anaconda3\lib\site-packages\keras\utils\data_utils.py", line 661, in <lambda> initargs=(seqs, self.random_seed)) File "C:\ProgramData\Anaconda3\lib\multiprocessing\context.py", line 119, in Pool context=self.get_context()) File "C:\ProgramData\Anaconda3\lib\multiprocessing\pool.py", line 176, in __init__ self._repopulate_pool() File "C:\ProgramData\Anaconda3\lib\multiprocessing\pool.py", line 241, in _repopulate_pool w.start() File "C:\ProgramData\Anaconda3\lib\multiprocessing\process.py", line 112, in start self._popen = self._Popen(self) File "C:\ProgramData\Anaconda3\lib\multiprocessing\context.py", line 322, in _Popen return Popen(process_obj) File "C:\ProgramData\Anaconda3\lib\multiprocessing\popen_spawn_win32.py", line 89, in __init__ reduction.dump(process_obj, to_child) File "C:\ProgramData\Anaconda3\lib\multiprocessing\reduction.py", line 60, in dump ForkingPickler(file, protocol).dump(obj) TypeError: can't pickle generator objects

Relevant warnings:
INFO:numexpr.utils:Note: NumExpr detected 16 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 8. INFO:numexpr.utils:NumExpr defaulting to 8 threads.

Environment info:

• Windows 10
• keras==2.2.4
• tensorflow==1.14
• Processor: intel(R) core(tm) i9‑9900k CPU @ 3.10 ghz

Automatic log folder creation for UNet does not work

When running InstantDL for Semantic Segmentation, the logs folder is not created automatically in ".../data-directory/logs". The program fails after 1 epoch since the weights can't be saved.

testing with pre-trained model

Hello,

I'm interested in testing out the InstantDL workflow using images of fluorescent nuclei from the 2018 DSB stage 1 test dataset (https://bbbc.broadinstitute.org/BBBC038) before training my own model. I'd love to try out the model that you've trained on this data for your paper on biorxiv (Fig. 3A). Is it possible to do so with the current InstantDL workflow?

Thanks! Looks like a cool resource and I'm excited to check it out.
Pearl

Malis Loss default settings

As Malis Loss can only be used with batch size = 1 this should be the default setting overwriting other batch sizes set by the user when using Malis Loss (Malis branch).

Logs folder in UNet

The /logs folder seems not to be created in the right directory if used the first time with the UNet

Tensorflow Version

Version mismatch between Readme and setup

InstanceSegmentation import of data_augmentation

I get a ModuleNotFoundError: No module named 'data_generator' when running RCNNmodel.load_image_gt.

Import is done as (here):
from data_generator.data_augmentation import data_augentation

Shouldn't it be:
from instantdl.data_generator.data_augmentation import data_augentation ?