Questions about one function/arguement

[Gooda11]

I try to use meshlib.mrmeshpy.Vector3f(520/7440, 40/350, 50/150) to create 3d model from .tiff files. For each .tiff slice, x range is(-260,260) and y(-20,20). I have 150 tiff files and the total length is 50. I try to make real scale model by using the argument. However, the model has double length in y direction than real scale. For tiff slice, I used 'plt.savefig(filepath, dpi=1200,
transparent=False, bbox_inches='tight',pad_inches = -0.1,)' .
I don't know what is wrong. The length in z and x direction is ok but y length is doubled.

[Grantim]

Hello! Could you please describe what is 7440 and 350 in your voxel size? (I thing X and Y components of voxel size should be equal in most cases)

[Gooda11]

Hello, 7440 is x_pixel and 350 is y_pixel. My tiff file is of 7440*350 dimensions, as the screenshot shows

[Gooda11]

This is the screenshot I want to attach! Sorry for confusion!

[Grantim]

In this case calculations looks legal. How did you find this range x(-260,260) and y(-20,20), if you could share you files (you can send them on support@meshinspector.com if you dont want to reveal them in public) we could investigate further

[Gooda11]

I set this range to be like that when saving the tiff files. Ok, I will send a zip file to this email via google email. Thanks!

[Grantim]

Closing, since this was followed up by email. A summary for anyone who finds this thread:

Voxel size. When loading a TIFF stack, every pixel becomes one voxel, and voxelSize is the real-world size of one voxel along each axis. Pixels are square in almost all cases, so voxelSize.x and voxelSize.y should normally be equal. Compute the value from the axis whose physical range you know reliably and use it for both:

voxelSize = ( physical X range / image width in px,
              physical X range / image width in px,
              stack length     / number of slices )

So here it should be (520/7440, 520/7440, 50/150). With 40/350 for Y, the Y voxel is about 1.6× larger than the X voxel (0.114 vs 0.070), and that is the stretched Y axis: the images didn't actually cover exactly 40 units of Y over their 350 px. To check: volume.dims × voxelSize should equal the physical size you expect.

Iso value can be anything between the volume's min and max. It is not limited to 0–1, and not always 0–255 either, since TIFF can store different value types.

No TIFF step needed. If the data is already a NumPy array, you can build the volume from it directly with mrmeshnumpy.simpleVolumeFrom3Darray(arr) (float32/float64; min/max are filled in automatically), then set voxelSize. See test_python/test_voxels_conversion.py.