Suggestions on setting up Bisque system on our own server

[rav...@ucsb.edu]

Hello,

I'm a grad student at UCSB, advised by Kelly Caylor at the Earth Research Institute and Department of Geography and also working with Tyler Clavelle of the Sustainable Fisheries Group, also at UCSB. We are interested in setting up Bisque on our ERI server to label satellite images. 

Our labeling tasks will involve digitizing 100s of tiff images, each of which is 150 megabytes. We also would like the ability to toggle between two different spectral band combinations, true color and false color infrared. I was curious if you had any recommendations on how much memory our server should have to quickly display, pan, and zoom on these large images. I tested out the dockerized Bisque on my laptop with 8gb of RAM and it crashed. I've also tested out a 150 Mb image on the Bisque served from https://bisque.ece.ucsb.edu/ and it also appears to error, only rendering the left most part of the image. One of our servers at ERI has 126 Gb of RAM (which I assume is the limiting factor), would this be enough to seamlessly label large images? or is there a cap to the size of the image that Bisque can support? We have terabytes of storage available, 8 cores, and two 1080 Ti GPUs, in case any of that matters.

I'm also curious if there's any bandwidth amongst your team to help our ERI IT staff install the dockerized Bisque on one of our servers. We seem to be having trouble with allowing users to access the ports due to some trickery with our network and docker.

Any advice and help is appreciated!

Cheers,

Ryan

[kgk]

A 150MB image should be no problem, so it it is something else.  If you could share (privately) the image uri with me (k...@ucsb.edu), I will see if I can check the error.

We provide dockerized image for testing, but these are usually customized for real use (filesystem, external DB, etc)..  I can help next week with suggestions for your setup.

--
You received this message because you are subscribed to the Google Groups "bisque bioimage" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bisque-bioima...@googlegroups.com.
To post to this group, send email to bisque-...@googlegroups.com.
Visit this group at https://groups.google.com/group/bisque-bioimage.
For more options, visit https://groups.google.com/d/optout.

--

Kristian G. Kvilekval (kkvil...@gmail.com) 805-636-1599

[Dmitry Fedorov]

Hi Ryan,

Why to necessarily install and maintain your own version of BisQue? CyVerse (U Arizona) <https://www.cyverse.org/> is running a large infrastructure and BisQue is one of the products they are offering for researchers. It might be much easier to simply use an existing deployment.

-dmitry

--

You received this message because you are subscribed to the Google Groups "bisque bioimage" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bisque-bioima...@googlegroups.com.
To post to this group, send email to bisque-...@googlegroups.com.
Visit this group at https://groups.google.com/group/bisque-bioimage.
For more options, visit https://groups.google.com/d/optout.

--

__________________________________

Dmitry Fedorov Levit <di...@dimin.net>
Web: http://www.dimin.net/
__________________________________

[rav...@ucsb.edu]

Hi Dmitry,

Thanks for the advice, we are going to test out Cyverse. We spoke to Kristian about the COG support issue, here are some articles that might be helpful in getting Bisque to support this increasingly common format for geospatial imagery: https://medium.com/planet-stories/a-handy-introduction-to-cloud-optimized-geotiffs-1f2c9e716ec3

Some new users are NASA, DigitalGlobe, and Planet: http://www.cogeo.org/ We are using Planet imagery for our project.

Cheers,

Ryan

[Dmitry Fedorov]

Hi Ryan,

I’ve already had requests to support COGEO, not sure why it’s not being red right away since the format suggests tiff compatibility. I’ll take a look at it when I have time.

It would be helpful if you could tell me a bit about your project, goals and how you are planning to use bisque. For example, what are you labeling and are you interested in training an automated system for this purpose.

Image size has no limits in bisque but large images should be written in a tiled pyramidal format otherwise a server with enough ram will be needed to further convert the images into a supported format. I need to double check that cogeo actually supports that. In any case it is always possible to convert images into one already supported format prior to upload.

-Dmitry