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.

Overlapping patches for training

See original GitHub issue

Hey,

Is there any way to obtain overlapping patches for training? I was thinking something similar to https://torchio.readthedocs.io/data/patch_inference.html#torchio.data.GridSampler but with the overlap parameter enabled.

Cheers, Sarthak

Issue Analytics

State:
Created 3 years ago
Comments:5 (5 by maintainers)

Top GitHub Comments

1reaction

fepegarcommented, Oct 27, 2020

I’m going to close this issue, as it’s basically the same as #341. We can continue the discussion there.

0reactions

sarthakpaticommented, Oct 27, 2020

What do you mean by “patch construction”?

I meant when we use one of the patch samplers.

This can be achieved (during inference) with thepatch_overlap kwarg of the GridSampler.

Yes, I know this (I have referenced this in my first post). I am asking if this can be done with the training sampler as well.

Top Results From Across the Web

Segmentation using overlapping patch extraction in ...

We used large overlapping image patches as inputs for intrinsic and extrinsic data augmentation, a patch selection algorithm, and a patch prediction fusion ......

Slicing images into overlapping patches at runtime

An alternative solution is to divide each image into overlapping slices, and using these during training and inference, before aggregating the predictions; ...

Learning One Convolutional Layer with Overlapping Patches

We give the first provably efficient algorithm for learning a one hidden layer convolutional network with respect to a general class of (potentially...

Creating training patches for Deep Learning Image ... - Medium

The patches will be created automatically and exported as a collection of images. Normally there is no overlapping between patches, but one may ......

Patch-Based Generative Adversarial Neural Network Models ...

To test effects of overlapping sCT patches on estimations, we 1) trained the models for three orthogonal views to observe the effects of ......