cvg / Hierarchical-Localization

Visual localization made easy with hloc
Apache License 2.0
3.24k stars 599 forks source link

Sift based local feature extraction is not working bcause of update in pycolmap. #370

Open dev111ce opened 9 months ago

dev111ce commented 9 months ago

The local feature extractor SIFT based DOG extractor class model in hloc/extractors/dog.py seems to be now broken becuase of the update in the pycolmap. With the latest pycolmap 0.6.1, pycolmap.sift.extract returns only two parameters while three are expected in the current code. keypoints, scores, descriptors = self.sift.extract(image_np). Upon execution, we get the below error:

ValueError: not enough values to unpack (expected 3, got 2)

Upon performing a quick RCA, we observed that the current pycolmap i.e., 0.6.1 version, returns just keypoints and descriptors and omits scores, which is the reason for the breakdown.

Any suggestion/fix on how to obtain the score value or fix the existing solution is greatly appreciated. Thanks in advance.

sarlinpe commented 9 months ago

This has been fixed in https://github.com/cvg/Hierarchical-Localization/commit/936040e8d67244cc6c8c9d1667701f3ce87bf075, please update your clone of hloc.

dev111ce commented 8 months ago

I am installing the latest pycolmap 0.6.1, so I have to go back to 0.6.0?