Cars

Added

- New default pipeline sensors_to_dense_dsm_no_merging with cumulative rasterisation [698]
- Profiling report for multiprocessing mode [745]
- Dashboard showing tiles processing during a run [765]
- Save of sparse matches in sensor geometry [761]
- Relaunch of frozen tasks after timeout [768]

Changed

- Version 1.6.0 of Pandora is used [688]
- Version 0.2.0a2 of Shareloc is used [756]
- Sparse matching processed by strip to improve computation time [753]
- DEM generation improved : less artifacts, filling of invalid regions [754]
- Default orchestrator is multiprocessing [755]
- Disparity interval cropped when it is too high [757]

Fixed

- Data is typed to save memory consumption [661]
- Preload of CARS in workers for multiprocessing mode to reduce forking time [730]
- CPU usage limited to 100% per worker [750]
- Output classification pixels cannot be floating-point numbers anymore
- ROI not used for endogeneous DEM [772]

Added

- Local disparity ranges are used [695]
- Write point cloud by pair [727]
- Left and right classification fusion on disparity map [729]

Changed

- Plane filling method: no propagation by default [725]

Fixed

- Slurm cluster launch [736]
- None tiles managed in save_cars_dataset [716]
- Color data type propagated to the end [642]
- Memory unreleased in multiprocessing mode [731]
- Loglevel could not be set [734]

Added

- CARS can now be pip installed [639]
- Name of color bands propagated in point cloud [696]

Changed

- Slurm account now as a parameter of Orchestrator [712]
- pkg_resources not used anymore [700]

Fixed

- Rasterization in epsg:4326 supported [724]
- Estimation of DSM roi in Shareloc plugin [701]

Added
- Multiprocessing cluster: add factorization of Delayed [673]
- 8 bits images are managed with particular sift parameters [668]
- Self calculated initial elevation with sift matches [664]
- Add debug mode with roi using initial tiling [475]
- Information map about dense matches filling [679]
- Script to generate CARS configuration [678]

Changed
- Multi points direct location [708]
- Application parameter "method" now optional [692]
- Geometry plugin interface [623]
- Nodata parameter now optional [296]

Fixed
- Orchestrator used does not match the one reported in the logs [681]
- used_conf is no longer created when the application is initialized [693]
- Parameter "mask" can be used in "dense_point_clouds_to_dense_dsm" [722]

Added
- Dask Slurm cluster [624]
- Crop disparity range with thresholds [626]
- Used source information map generated [648]

Changed
- Write workers logs in one single file [659]
- Use tempfile to generate default orchestrator folder instead of /tmp [663]

Fixed
- Cython upgrade bug [674]
- Too much used disk memory in MP mode [632]

Added
- Notebook: add masks and classification options [615]
- Option: Set python interpreter to use in PBS Dask Cluster [611]
- DensePointCloudToDenseDSM Pipeline: now re-entrance with confidence and classification [602]
- Shareloc Geometry plugin is now internal [618]
- Use_sec_disp option allowing to use right disparity map is removed [638]
- CARS can now be installed without OTB [637]
- Epipolar tiles corresponding to region outside sensor images are removed before any computation [635]
- Add CARS progress messages [658]

Changed
- Multiprocessing cluster doesn't freeze when bug occur anymore [607]
- Resampling is not done with OTB anymore [508]
- Pyproj is no longer constrained [646]

Fixed
- Notebook: retrieve dsm [614]
- Doc: Roi parameter disappearing [616]