Skip to content

Latest commit

 

History

History
61 lines (53 loc) · 3.82 KB

notes.md

File metadata and controls

61 lines (53 loc) · 3.82 KB

To Do List

Priority To Do

  • Improve speed, expecially for large datasets
  • Store matches at low resolution and use them in all the following steps when they are needed (without extracting them again)
  • Prepare pycolmap.yaml and calibration.yaml (for both colmap and openmvg)
  • Testing on very large datasets ([Issue #29])
  • Use Github submodules instead of copying thirdpary code inside the repo
  • Add subpixel refinement of the matches (e.g., cross-correlation or pixel-perfect-sfm)
  • Make semi-dense matcher work with multi-camera (Issue [#24])
  • Improve usage of multiple descriptors together
  • Finish extending compatibility to OpenMVG

Bugs and Issues

  • seems there is a bug in kornia when matching more than 1000 images
  • GUI is broken with the new configuration management
  • SE2-LoFTR not fully implemented
  • DeDoDe and RoMa are not working without CUDA
  • open issue on pydegensac for pairs with many tie points (>10000). Quite random results with LoFTR and RoMa
  • script run from scripts folder are not able to import modules!
  • scripts/delete_static_matches.py not ready
  • Avoid importing duplicated matches into colmap database when working with tiles bug (see Issue [#8])

Other enhancements

  • Improve configuration management Hydra to make using yaml files, command line and GUI (Issue [#48])
  • Tests on satellite images
  • Add steerers + DeDoDe
  • Add Silk features
  • Add SIFT + LightGlue
  • Support for more hand-crafted local features from kornia and opencv (e.g., affine SIFT, AKAZE, etc.)
  • Extend documentation
  • Workflow to rebuild & publish image to Docker Hub
  • Integrate support for Pix4D Open Photogrammetry Format
  • Improve GUI (e.g., image-matching-webui)
  • Improve visualization routines for extracted features and matches
  • Extend compatibility to other software (e.g., Meshroom)
  • Remove unmactched features from the h5 file and from COLMAP database after the matching is completed enhancement (Issue [#30])
  • Support for 16-bit images (Issue [#42])
  • Improve Tiler with adaptive tile grid computation enhancement (Issue [#23])
  • keep matching score/confidence (if available) (Issue [#22])
  • Allow to read images from different folders (e.g., subfolders)
  • Support for exporting matches in COLMAP database with different camera calibration based on the input images (e.g., on exif or on subfolder structure) (Issue [#45])
  • Apply masks during feature extraction
  • Automatically download weights for all the models
  • Autoselect tiling grid in order to fit images in GPU memory

Completed

  • Tile processing for high resolution images
  • Manage image rotations
  • Add image retrieval with global descriptors
  • add GUI
  • Add pycolmap compatibility
  • Add exporting to Bundler format ready for importing into Metashape (only on Linux and MacOS by using pycolmap)
  • Dockerization
  • Add tests, documentation and examples (e.g. colab, ..)
  • Cleanup repository to removed large files from Git history
  • Update README CLI options