[wip] OpenOrganelle demo

[kephale]

No description provided.

[kephale]

@tpietzsch I'm getting stuck on the OpenOrganelle demo because the N5 data is UnsignedByteType, but clearly it needs to get rendered as an UnsignedShortType.

However, the N5 is being opened as Volatile:

sciview/src/main/java/sc/iview/commands/demo/advanced/OpenOrganelleDemo.java

Line 246 in 5a3d0ea

final CachedCellImg<?, ?> vimg = N5Utils.openVolatile( n5, datasetsToOpen[s] );

scenery will need some corresponding updates for proper detection of multiscale images that are opened with N5 (currently it is focused on detecting normal BDV images with XML metadata).

[tpietzsch]

@kephale
First, note that N5Utils.openVolatile will still open the image as a (non-volatile) UnsignedByteType image. (But with underlying VolatileByteAccess that provides the capability to turn it into a volatile image.)

At some point you need to use VolatileViews.wrapAsVolatile(...) to make it volatile.
I wish I could tell you that you can convert it to UnsignedShortType and it will just work. Unfortunately it wont, yet, because VolatileViews can't deal with the Converter.

You could try to first VolatileViews.wrapAsVolatile(...) to turn it into RandomAccessibleInterval<UnsignedByteType>. And then convert from VolatileUnsignedByteType to VolatileUnsignedShortType. I fear that it will still not work because BVV assumes that it will get some kind of AbstractCellImg, and the converted RAI will not be that.

So, unfortunately you'll have to make a (lazy) copy... :-(
I'm pretty sure this will work:

import bdv.util.volatiles.VolatileViews;
import net.imglib2.RandomAccessibleInterval;
import net.imglib2.algorithm.blocks.BlockAlgoUtils;
import net.imglib2.algorithm.blocks.convert.Convert;
import net.imglib2.blocks.PrimitiveBlocks;
import net.imglib2.cache.img.CachedCellImg;
import net.imglib2.type.numeric.integer.UnsignedByteType;
import net.imglib2.type.numeric.integer.UnsignedShortType;
import net.imglib2.type.volatiles.VolatileUnsignedShortType;

static RandomAccessibleInterval< VolatileUnsignedShortType > convert(
		RandomAccessibleInterval< UnsignedByteType > rai )
{
	final int[] cellDimensions = { 64, 64, 64 };
	final PrimitiveBlocks< UnsignedByteType > blocks = PrimitiveBlocks.of( rai );
	final CachedCellImg< UnsignedShortType, ? > img = BlockAlgoUtils.cellImg(
			blocks,
			Convert.convert( new UnsignedByteType(), new UnsignedShortType() ),
			new UnsignedShortType(),
			rai.dimensionsAsLongArray(),
			cellDimensions
	);
	return VolatileViews.wrapAsVolatile( img );
}

It requires recently released versions

<net.imglib2.imglib2.version>6.2.0</net.imglib2.imglib2.version>
<net.imglib2.imglib2-algorithm.version>0.14.0</net.imglib2.imglib2-algorithm.version>

Of course, the better solution would be to make BVV just handle these datatypes, but I don't have a timeline for that.

[kephale]

Thank you very much @tpietzsch ! You've unblocked the demo work 🍪 !

[kephale]

This is getting stuck here https://github.com/scenerygraphics/scenery/blob/be284d2488c8635952d0a62f5ca3e5bed15742d1/src/main/kotlin/graphics/scenery/volumes/VolumeManager.kt#L396. it seems like it is trying to create a ton of FillTasks (129705 blocks?), the next thing i would check is how the BVV volume is being initialized w.r.t. the camera

[tpietzsch]

That sounds like it doesn't use multi-resolution

[kephale]

migrated to #537