[HN Gopher] Spectral Imaging Made Easy: A Powerful Python Library
___________________________________________________________________
Spectral Imaging Made Easy: A Powerful Python Library
Author : janezla
Score : 55 points
Date : 2024-12-23 11:39 UTC (2 days ago)
(HTM) web link (github.com)
(TXT) w3m dump (github.com)
| janezla wrote:
| I've created a Python library for working with spectral images.
| It started as a mix of work and personal interest. Since I work
| in research, I brought together a lot of useful code to make
| handling spectral images easier and packaged it into this
| library. I hope others find it helpful too! :blush:
|
| Link to docs: https://siapy.github.io/siapy-lib/
| fooblaster wrote:
| what exactly does one do with hyperspectral images? Or what do
| you do with your library?
| boccaff wrote:
| There is a multitude of applications leveraging parts of the
| spectra different than the visible. I come from an
| agricultural background, and you can see examples from
| improving classification of land use, detection and
| classification of diseases, nutritional status assessment,
| indirect measurements of properties of plants and soil... it
| is endless, and every time any part of the tool stack gets
| cheaper, you have more and more potential applications. This
| comment [1] have a nice description for the library.
|
| [1] https://news.ycombinator.com/item?id=42507805
| __mharrison__ wrote:
| Hint. If your library is for creating images... Put an example
| image in the Readme.
| tweakimp wrote:
| I tried to understand what this library does, but without image
| examples its impossible for me. The docs almost seem to be
| unhelpful on purpose. Look at the use case description: "The
| functionality of the SiaPy library has been implemented in
| various use cases, demonstrating its capabilities and potential
| applications. The library's functionality is not limited to
| these examples and can be extended to other applications as
| well."
|
| Are we living in the dead Internet already where everything is
| meaningless AI garbage?
| Davidbrcz wrote:
| Spectral images are images where there are several sensors
| into one image (think visible and infrared/thermal for
| instance). A good example would be Altum Pt camera
| (https://ageagle.com/drone-sensors/altum-pt-camera)
|
| Then, this library can be used for instance (their word) -
| Display images from two cameras. - Co-register cameras and
| compute the transformation from one camera's space to
| another. - Select regions in images for training machine
| learning (ML) models. - Perform image segmentation using a
| pre-trained ML model. - Convert radiance images to
| reflectance by utilizing a reference panel. - Display
| spectral signatures for in-depth analysis.
| toxik wrote:
| It is a by-product of a research project, its main connection
| is "these things were useful to the author while working on
| spectral images".
| ulrischa wrote:
| I made spectral image analysis at university. And there weren't
| good software Tools available
| hoomanmo wrote:
| is it compatible with Python 3.13?
| KeplerBoy wrote:
| Isn't pretty much everything compatible with 3.13?
|
| The packages, which were affected by breaking changes (numpy,
| cython, scipy and so on) were patched months ago.
| mturmon wrote:
| Related: A python package for atmospheric correction of imaging
| spectroscopy ("hyperspectral") radiance data:
| https://github.com/isofit/isofit
|
| And a superset package, for the EMIT imaging spectroscopy
| investigation: https://github.com/emit-sds
| tomtom1337 wrote:
| If you're looking to analyse your hyperspectral images (spectrum-
| images, image-images or n-dimensional- n-dimensional datasets), I
| can highly recommend hyperspy [1].
|
| One of the brilliant ideas hyperspy incorporates is that we
| consider datasets to have a navigation dimension and a signal
| dimension (think, you measure a spectrum at each point on an
| image), and you can easily transpose between them. This means
| that you can <<move around>> on the image and see what the
| spectrum looks like, or transpose and see what the image looks
| like as a function of the spectrum.
|
| In particular I think the model building, where you can fit
| components to your dataset, is really useful.
|
| It works best with the Jedi LSP - pyright doesn't support the way
| we added lazy loading / extensions to the base hyperspy package.
|
| [1] https://hyperspy.org/
| ptero wrote:
| Hyperspy is great and the ability to "move around"
| n-dimensiobal datasets is a very powerful tool for the data
| visualization!
|
| When I used it I missed two things compared to a similar
| superpower tool I used when I was working with multidimensional
| field test data in Matlab.
|
| 1. Ability to use "text dimensions", or non-uniformly spaced
| grid points.
|
| 2. Ability to select and filter on arbitrary expressions
| instead of by slice only.
|
| The need for (2) is harder to grok (what's that going to do for
| a grid dataset???), but being able to apply a few arbitrary
| selection expressions is a superpower when analyzing messy 10+
| dimensional data.
|
| That, and the ability to add, on the fly, virtual dimensions
| for arbitrary expressions.
|
| Someday, when I am ready to retire, I will take half a year to
| build this in python...
| tomnicholas1 wrote:
| Interesting - I'm curious whether you feel that Xarray covers
| these use cases already?
|
| https://xarray.dev/
|
| Especially as I've said before that Hyperspy shares so many
| features in common with Xarray that Hyperspy should just use
| Xarray under the hood.
|
| https://github.com/hyperspy/hyperspy/discussions/3405
| adammarples wrote:
| All that work and you can't put a description of what it does, an
| example, an image, something. 10'000 people click the link you
| posted, see nothing at all, and leave again.
| ipunchghosts wrote:
| https://github.com/isaacgerg/matlabHyperspectralToolbox
| BugsJustFindMe wrote:
| I spent 20 minutes clicking through links and reading
| descriptions and I still can't tell whether this is for pictures
| of ghosts or something else.
| momoschili wrote:
| who out there actually has a consumer spectral imager these days?
| Cheapest ones I can find are ~10k USD....
___________________________________________________________________
(page generated 2024-12-25 23:01 UTC)