Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Transforms enhancement for HoVerNet

See original GitHub issue

In the original HoVerNet repo(https://github.com/vqdang/hover_net), they use imgaug(https://github.com/aleju/imgaug) to do data augmentation, which has several differences with MONAI.

add new_key in Lambda (addressed by #5526) Now in Lambdad, if we set overwrite to False, it will not return data after the Lambda function. We may add an argument called new_key, then if it is not None, we can get the data after the Lambda function. And If it is None with overwrite set as False, we just get the behavior as before. https://github.com/Project-MONAI/MONAI/blob/5332187324bc2471476581e729d685e30d03d40a/monai/transforms/utility/dictionary.py#L1035
add RandShiftSaturation and RandShiftHue This one can be replaced by TorchVisiond this one. https://github.com/Project-MONAI/MONAI/blob/5332187324bc2471476581e729d685e30d03d40a/monai/transforms/utility/dictionary.py#L1370 https://github.com/Project-MONAI/tutorials/blob/f43e771e7cf35427715baa16245e8a12dffe423a/pathology/tumor_detection/ignite/camelyon_train_evaluate.py#L89-L91
add RandomOrder (addressed by #5534) When we have a list of transforms, we may want these transforms to do in random order. We can get a reference from imgaug repo. https://github.com/aleju/imgaug/blob/0101108d4fed06bc5056c4a03e2bcb0216dac326/imgaug/augmenters/meta.py#L3015-L3017

We may need to enhance our transforms to align with their implementation. cc @wyli @Nic-Ma

Issue Analytics

State:
Created 10 months ago
Reactions:1
Comments:5 (5 by maintainers)

Top GitHub Comments

1reaction

Nic-Macommented, Nov 16, 2022

Hi @KumoLiu ,

Please don’t mark “fix # XXX” in a PR for this ticket, and click the correct checkbox when PR merged.

Thanks.

1reaction

wylicommented, Nov 14, 2022

Could you please put more details in the feature requests？ these are quite confusing to me.

Top Results From Across the Web

Hover-Net: Simultaneous segmentation and classification of ...

In this paper, we present a deep learning approach 3 for simultaneous segmentation and classification of nuclear instances in histology images. The network...

[2203.01940] Color Space-based HoVer-Net for Nuclei ... - arXiv

Secondly, to enhance the visual differences between nuclei, a multi-channel color space-based approach is used to aid the model in extracting ...

HoVer-Net: Simultaneous Segmentation and Classification of ...

The development of automated methods for nuclear segmentation and classification enables the quantitative analysis of tens of thousands of ...

Separable-HoverNet and Instance-YOLO for Colon Nuclei ...

In this work, we propose an approach that combine Separable-HoverNet and Instance-YOLOv5 to indentify colon nuclei small and unbalanced.

Releases · Project-MONAI/MONAI - GitHub

Hover-Net based digital pathology workflows including new network, loss, postprocessing, ... Enhancements for NVTX Range transform logging; Enhancements for ...