Paul-Edouard Sarlin

So far we've been relying on the geometry objects available in COLMAP. We could consider adding support for Sophus objects if there is a need for minimal dependencies within pyceres.

Nice, thanks for sharing! For larger problems, you would likely need to pre-compute the forward and backward passes for all residuals and parameters in parallel (e.g. in an evaluation callback)...

Depending on where the glog library was installed, you might need to make it available, for example by running this in bash: ```bash export LD_LIBRARY_PATH=/usr/local/lib:${LD_LIBRARY_PATH} ```

As mentioned in the pycolmap README, the development has moved to [the colmap repo](https://github.com/colmap/colmap/tree/main/pycolmap). The cost functions are in [src/pycolmap/estimators/cost_functions.h](https://github.com/colmap/colmap/blob/main/src/pycolmap/estimators/cost_functions.h) but haven't made it into the previous release, so you...

You likely need to install pyceres from source as well. > I may not install the pycolmap properly, as I copied colmap src folder to pycolmap subfolder and changed the...

> Thanks for the help. I tried to install pyceres from sources. Now there seems to be a conflict between two libraries pycolmap and pyceres. When they are imported together,...

Were pyceres and pycolmap both installed from source? This this should not occur. Did it regardless of the order in which you import them?

Hey @ducha-aiki, We've now released the LightGlue models trained for ALIKED & SIFT: https://github.com/cvg/LightGlue/commit/29f3e449efa1994758b8a16299d2816028dca65b These models are frankly very good - we've added [results on MegaDepth](https://github.com/cvg/glue-factory#megadepth-1500) and will add results...

- kornia is easier to install than pycolmap (GPU-SIFT requires building from source, we don't yet have the bandwidth to ship cuda-enabled wheels) - when considering batching, a torch-SIFT could...

We figured out later that we could simplify the implementation by training the confidence classifier at the same time as the rest of the model but detaching its inputs. This...