Batch size > 1 images with different dimensions

[jaysnanavati]

Currently this is not possible, any pointers on how to work around this?

[roman-kh]

You need to create a custom batch class which stores an array of arrays with images.

[jaysnanavati]

hmm how would that work with the other functions that work on just a single array of images?

[roman-kh]

That would require some heavy refactoring. But overall in the long run this will simplify the code and the data structure.

[jaysnanavati]

Indeed, though what if we add the ability to pass in the same parameters as unifysampling to load? that way as soon as the image is loaded we can call unify to make sure all of them are the same dims?

---

From: Roman Kh notifications@github.com Sent: Wednesday, April 24, 2019 11:53:34 PM To: analysiscenter/radio Cc: Jay Nanavati; Author Subject: Re: [analysiscenter/radio] Batch size > 1 images with different dimensions (#38)

That would require some heavy refactoring. But overall in the long run this will simplify the code and the data structure.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/analysiscenter/radio/issues/38#issuecomment-486457195, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AAURCSKJFYSNSU5S7JT6VMLPSDQG5ANCNFSM4HH3POSQ.

[roman-kh]

If you want just load images with different dimensions all you may need is a custom load action, which load images as is and then reshapes them to a common shape.

[jaysnanavati]

Yes, but I'd like to reuse existing implemented functions to do this :) what are your thoughts on having these parameters in the load function ?

---

From: Roman Kh notifications@github.com Sent: Thursday, April 25, 2019 12:33:19 PM To: analysiscenter/radio Cc: Jay Nanavati; Author Subject: Re: [analysiscenter/radio] Batch size > 1 images with different dimensions (#38)

If you want just load images with different dimensions all you may need is a custom load action, which load images as is and then reshapes them to a common shape.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/analysiscenter/radio/issues/38#issuecomment-486635422, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AAURCSL7ERTBJ5MBVTFHWCDPSGJH7ANCNFSM4HH3POSQ.

[jaysnanavati]

Ping??

[roman-kh]

Make a PR, please, and we will discuss your implementation.

[jaysnanavati]

Awesome! thanks for your help on this, appreciate it :)