jefferislab / NBLAST_on-the-fly

Web application demonstrating NBLAST neuron similarity/search
http://jefferislab.org/si/nblast
5 stars 3 forks source link

Allow input of .NRRD file format #65

Closed josesho closed 9 years ago

josesho commented 9 years ago

Allows upload of skeletonised neuron in .NRRD file format via dotprops function

jefferis commented 9 years ago

Dear @josesho Joses,

Many thanks for your contribution. Very good idea!

I added various kinds of checks and documentation, and the patch is now merged to master.

Best,

Greg.

josesho commented 9 years ago

Hi Greg,

No problem, thanks for accepting the pull.

I tried uploading a .NRRD file of my own (Downsampled to 280 x 160 microns, file size 48KB) , but got this error instead "cannot allocate vector of size 187.0 Mb". Sounds like a restriction on your server side?

Joses

On 18 August 2015 at 04:01, Gregory Jefferis notifications@github.com wrote:

Dear @josesho https://github.com/josesho Jose,

Many thanks for your contribution. Very good idea!

I added various kinds of checks and documentation, and the patch is now merged to master.

Best,

Greg.

— Reply to this email directly or view it on GitHub https://github.com/jefferislab/NBLAST_on-the-fly/pull/65#issuecomment-131944431 .

jefferis commented 9 years ago

The VPS is quite small and I also ran into this once. Have your tried a few times?

But if you raise an issue, I've thought of a way to reduce the memory usage further. Best,

Greg.

Sent from my iPhone

On 18 Aug 2015, at 04:01, Joses W. Ho notifications@github.com wrote:

Hi Greg,

No problem, thanks for accepting the pull.

I tried uploading a .NRRD file of my own (Downsampled to 280 x 160 microns, file size 48KB) , but got this error instead "cannot allocate vector of size 187.0 Mb". Sounds like a restriction on your server side?

Joses

On 18 August 2015 at 04:01, Gregory Jefferis notifications@github.com wrote:

Dear @josesho https://github.com/josesho Jose,

Many thanks for your contribution. Very good idea!

I added various kinds of checks and documentation, and the patch is now merged to master.

Best,

Greg.

— Reply to this email directly or view it on GitHub https://github.com/jefferislab/NBLAST_on-the-fly/pull/65#issuecomment-131944431 .

— Reply to this email directly or view it on GitHub.

jefferis commented 9 years ago

re this:

I tried uploading a .NRRD file of my own (Downsampled to 280 x 160 microns, file size 48KB) , but got this error instead "cannot allocate vector of size 187.0 Mb". Sounds like a restriction on your server side?

Just a couple comments: First, it is the uncompressed size that matters, and that is probably quite a bit bigger.

Also out of curiosity, if the image is downsampled then it should be the number of pixels rather than the number of microns that changes. Were you cropping (which is fine, of course if you change the origin appropriately, something that ImageJ/Fiji doesn't handle)?

josesho commented 9 years ago

Yes, uncompressed size is on the order of 100 MB. I am downsampling the compressed file, and not the uncompressed version.

Secondly, both pixels and microns have changed, curiously, with no cropping whatsoever.

Original --> 564.25 X 326.71 microns, 1769 X 1026 pixels Downsampled --> 282.29 X 163.35 microns, 885 X 513 pixels

I simply use Plugins > Example > downsample. Or should I go about it a different way?

josesho commented 9 years ago

You are right, downsampling in Fiji alters the origin of the skeletonised neuron. (See below.) Better discuss this on the nat-users Google groups?

screen shot 2015-08-19 at 3 03 45 pm

jefferis commented 9 years ago

OK, see https://groups.google.com/forum/#!topic/nat-user/qQo44Qi1TkM for my recommended pre-processing steps. Basically you cannot reliably crop in Fiji and you should use Image ... Scale not Plugins ... Examples ... Downsample since the latter does not set voxel size correctly (and only scales the XY dimensions).