v0.6.0 - 2025-01-02
-
due to
24fe2d4
- split_df with file names instead of paths, required img_dir (commit by @martibosch):split_df with file names instead of paths, required img_dir
24fe2d4
- split_df with file names instead of paths, required img_dir (commit by @martibosch)
9657e62
- conda install lightgbm to avoid macos issues; use tox-gh (commit by @martibosch)0941763
- revert to tox.ini (see github.com/tox-dev/tox/issues/3457) (commit by @martibosch)
v0.5.0 - 2024-03-28
6bdbdc3
- rasterize_lidar with shape/transform args and return zeros (commit by @martibosch)c619239
- adaboost->generic classifier (default lgb), joblib->skops (commit by @martibosch)853d9ad
- fix to enforce all keyword-only args (commit by @martibosch)0729a3d
- accept pca/kmeans kwargs in train/test split (commit by @martibosch)a3b1362
- update to new ndi namespace for rotate (commit by @martibosch)4f50349
- pre-trained model from huggingface hub (commit by @martibosch)
bbb7c19
- l2c meta height/width instead of shape, lazrs lidar backend (commit by @martibosch)
694966a
- using opencv for faster convolution (commit by @martibosch)5c00c95
- droppedDEFAULT
from settings, consistent docs (commit by @martibosch)0c9da36
- kws->kwargs (commit by @martibosch)73ab0c4
- changeclassify_img
->predict_img
(commit by @martibosch)
9696124
- _create_tmp_dir to rm test tmp dirs if existing (commit by @martibosch)
- moved
_gaussain_kernel1d
from scipy to detectree codebase
- added postprocess func args and kwargs to
LidarToCanopy
- added pre-commit
- updated to laspy 2.0.0 (with optional laszip)
- updated bumpversion to double quotes (Python black)
- updated docs build
- tests and release to pypi and github with github actions
- added github issue templates, pull request template and updated contributing docs
- using pydocstyle and black
- added lidar to canopy module
- using keyword-only arguments
- drop
num_blocks
argument ofcompute_image_descriptor
andcompute_image_descriptor_from_filepath
- set default post-classification refinement parameter
refine_beta
to 50 (instead of 100) - keyword arguments to
PixelFeaturesBuilder
andPixelResponseBuilder
can be explicitly provided to the initialization ofClassifierTrainer
, and are documented there - raise a
ValueError
when a provided response is not a binary tree/non-tree image
- correction (typo)
keep_emtpy_tiles
->keep_empty_tiles
insplit_into_tiles
- initial release