Currently, we need to load in separate volumes for large microscopy datasets, where the volume to mesh operators only work on a single mesh. So a requirement is some form of operator for making multiple, linked meshes with volume to mesh operators.

This might be easiest by adding the same GN modifier to all linked meshes, and then doing some magic with volume to mesh nodes in there. this would need to be tested, though.

copying over into the right subfolder of Contents worked to fix this - find out why this didnt work

This happens in approx 5% of machines.

The node named "Material Output" is not created before it is used.
Thanks！

Python: Traceback (most recent call last):
File "C:\Users\Youming Xiao\AppData\Roaming\Blender Foundation\Blender\4.1\scripts\addons\tif2blender\ui\ui.py", line 75, in execute
load.load()
File "C:\Users\Youming Xiao\AppData\Roaming\Blender Foundation\Blender\4.1\scripts\addons\tif2blender\load.py", line 117, in load
vol_obj, volume_inputs = load_volume(volume_inputs, bbox_px, scale, cache_coll, base_coll, emission)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Youming Xiao\AppData\Roaming\Blender Foundation\Blender\4.1\scripts\addons\tif2blender\load_components\load_volume.py", line 142, in load_volume
materials = volume_materials(volume_inputs, emission_setting)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Youming Xiao\AppData\Roaming\Blender Foundation\Blender\4.1\scripts\addons\tif2blender\load_components\load_volume.py", line 106, in volume_materials
nodes.get("Material Output").location = (700,00)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'location'

right now it is always next to the images, which is not very easy to clean up

with shared shader and GN - for chunking; and driver for thresholding

time support is now done, guide people to how to change scale bars and what new import options do

Supporting extra image formats should be very doable, as the only thing this needs is a python library to import file to numpy array and reading metadata, the only issue in this is making Blender have optional dependencies for each of these files. And that then load and metadata functions need to be dynamically assigned based on file type.

To be clear, tif2blender would then still rewrite to (uncompressed) .vdb files, this is only about the input format into tif2blender.

Data types to support:

• OME-ZARR (ome-zarr-py)
• HDF5 (h5py), although this may have require some UI tuning because metadata names (even of the data) are not standardized
• LZW compressed .tif (imagecodecs with currently required tifffile) as noted in #35 by @hvdrosa
• .tif sequence (not the easiest to support, as this requires a file list)

Please comment with formats you'd want, or if you think you can help write a load script!

with maybe a force remake checkbox.

I think this slows down scrolling through time significantly, and unnecessarily

When I get to 4:57 of the tutorial there isn't a shader applied to the volume, as far as I can tell. Thanks for creating this great plugin!

Hi.
Just trying out to load some in vivo data from two photon microscopy (lower SNR than exM etc).
The volume loads (zxy) but no nodes are visible in the shading or geometry editors (screenshot attached).
Any clues?
Thanks!
P.

a lot more is possible there but some of the syntax (around group input/output for example) has changed.

Target FPS could probably be set if you do time stretching upon load
will be weird for multiple imports, but that can be a warning

Try to create a LUT-like interface for shaders by piping from white shader -> shader to RGB -> LUT and use Node tools in Blender 4 to create an Apply LUT.. interface.
How will this work for selecting channels?

have in import options a checkmark for 'mask':
this gives a channel selector which channel the mask is in
The mask can then be loaded as mesh (marching cubes per unique integer value)

In the scale bounding box geometry node I would add a node from "um per tick" to a "value to string" and then "string to curve" and "curve to mesh" so that the length corresponding to a single tick is displayed at 0,0,0

• auto threshold
• reload vdb
• set render settings

There is an inherent issue with scale bars for data visualization with perspective cameras, where a scale bar only works as long as the perspective can be accurately assessed by the human viewer, which can be hard with microscopy as normal landmarks do not exist.
options:

• simple axes of cylinders at 1 µm (simplest, minimal solution, asks the user to take care themselves)
• box enclosing the entire volume with ticks (gives perspective inference, the solution that fiji BVV does, however, this doesnt work with all camera placements)
• conversion factor of 1 blender meter to µm (or even setting the blender unit to physical µm, although this may be unfeasible). Communicating this may be hard, but an option could be adding a geometry nodes string-to-mesh of the conversion factor.

This could also all be optional inside a single GN node. I'm very open to other suggestions for this.

currently the first actions we do upon load are the same, adding Attribute > Map Range > Principled Volume with maybe Mix Shader of multiple for multiple channels, to make a first visualization of the data.

It would be nice to have this automatically upon load of a new tif, to also handle automatic linking of chunked volumes.

Should this automatically threshold? Maybe Otsu thresholding upon load would be good, and make you immediately capable of looking at your data.

otsu is slow, is nice to be able to turn off for large volumes

will be for a version 0.0.2 before moving to blender 4, together with tests created in #29, to have a 3.5/3.6 version of tif loader that is functional

As this is built on tifs, and fiji/imagej has such dominance, i think trying to read imagej_metadata for pixel size and axes order would be a nice QoL improvement.
This could be under a separate infer operator in the tif loader window

pyopenvdb is accessible if running blender from cli

Need to figure this out soon, has anybody else seen similar behaviour?
might be related to deprecation of pkg_resources API

do separate subgroup for crosshatch tick, selector for grid axes as inputs for scalebar group

In the current folders the vdb files are next to the tif, which is annoying for large volumes with chunked vdbs, might be nice to reintroduce a tmp folder for vdbs

and remove this from tick instance gen

vdb loading by frame is supported in blender, currently the addon only handles single frames. Might be good to move to a frame-by-frame folder with all chunks per frame.

Note: this may also change orders of chunk loading.

Hi Oane!

Amazing package! It also works with our cryoET data despite it being intrinsically noisy! (I had some minor issues with LZW compressed TIFs, but it worked nicely for non-compressed ones. I guess it's only a matter of some additional libraries I was missing to be able to read it).

I wanted to ask if you have implemented (or know) an easy solution to display and render only one selected slice (as one would do in the Data> Viewport display > Slice)

The issue is that whenever I go to render/shading mode, it displays the full volume and not only the selected slice (as in viewport)
.

Best,
Higor