Repository navigation
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Accumulate image-vector squared norms in at least float64. Currently the
einsumuses the image dtype, so ordinary int16 values fromread_bandscan overflow before the square root. This produces NaNs or incorrect angles even though the same spectra work when explicitly converted to float64, as discussed in #132.Use
np.result_typeto retain wider floating precision. The existing float64 endmember normalization and dot-product calculation stay the same.Add 14 scaled-direction cases covering signed/unsigned integers and floating types. All 16 spectral-angle tests pass on Python 3.8/NumPy 1.24 and Python 3.12/NumPy 2.5. The five numerical test files pass 165 tests, and the required Flake8 check passes.
On the official AVIRIS sample, all 12 dtype/read-path configurations now match an independent geometric angle reference and its nearest-endmember decisions. The three float64 controls remain byte-for-byte identical.