Merge pull request #129 from initze/version_bump

Version bump

CHANGELOG.md

@@ -1,124 +1,191 @@
 # Changelog
 
+## [0.10.2] - 2024-05-08
+
+### Added
+
+- re-added weights and biases to training module
+
+### Changed
+
+- bugfixes of postprocessing scripts
+- code and syntax cleanup
+
 ## [0.10.1] - 2024-04-25
+
 ### Changed
+
 - added gdal as pre-requirement to readme
 - fixed path issue for postprocessing scripts
 
 ## [0.10.0] - 2024-04-24
+
 ### Added
+
 - pip-installable packaging
 
 ### Changed
+
 - entire project structure (to make it pip-installable)
 
 ## [0.8.0] - 2022-09-09
+
 ### Added
+
 - added Unet3+ model
 - added deep supervision loss plot
 
 ## [0.7.1] - 2022-08-02
+
 ### Added
+
 - added learning rate scheduler (StepLR, ExponentialLR)
 - added example configuration file
 
 ## [0.7.0] - 2022-03-29
+
 ### Added
+
 - added singularity container definition
 
 ## [0.6.1] - 2022-03-22
+
 ### Changed
+
 - fixed bug (crashed prepare_data.py for Planet Scenes)
 
 ## [0.6.0] - 2021-12-21
+
 ### Added
+
 - data path selection for all scripts
 - inference path selection for inference.py
+
 ### Changed
+
 - overhauled data structure: disentangled from processing/code directory
 - fixed masking bug (did not recognize udm values > 1) for PSOrthoTile data
 
 ## [0.5.5] - 2021-11-25
+
 ### Changed
+
 - added support for Multi- and single-GPU trained models
 
 ## [0.5.4] - 2021-11-16
+
 ### Changed
+
 - added Multi-GPU support for Training
 
 ## [0.5.3] - 2021-11-10
+
 ### Changed
+
 - added support for PSOrthotile input (flexible input data resolution)
 
 ## [0.5.2] - 2021-08-18
+
 ### Changed
+
 - fixed inference masking bug
 
 ## [0.5.1] - 2021-06-10
+
 ### Changed
+
 - bugs from 0.5.0 fixed
 - logfiles moved to logs directory
 - completely moved to argparse (from docopt) for argument parsing
 - (false) nodata removed from inference vector output
 
 ## [0.5.0] - 2021-05-19 - Broken Version
+
 ### Added
+
 - Training/Inference/Data Preparation logs are now written to corresponding log files
 - gdal configuration file
 - minor code cleanup
 
 ## [0.4.2] - 2021-03-19
+
 ### Added
+
 - shapefiles as inference output
+
 ### Changed
+
 - improved inference control figures
 
 ## [0.4.1] - 2021-03-11
+
 ### Changed
+
 - fixed broken inference
 
 ## [0.4.0] - 2021-03-01
+
 ### Added
+
 - selection of different DL models and backbones
 - implementation of augmentation types based on the albumentations package
 
 ## [0.3.3] - 2021-02-22
+
 ### Added
+
 - added robust futureproof datasource sorting
 
 ## [0.3.2] - 2021-02-15
+
 ### Added
+
 - added parallel preprocessing (setup_raw_data.py, prepare_data.py)
+
 ### Changed
+
 - changed option parsing from docopt to argparse for preprocessing scripts
 
 ## [0.3.1] - 2021-02-08
+
 ### Changed
+
 - fixed bug causing crash running inference.py
 
 ## [0.3.0] - 2021-01-26
+
 ### Added
+
 - Named training runs
+
 ### Changed
+
 - Loading Arctic DEM elevations and slopes locally
   - fixed too low resolution bug (loading from Google Earthengine)
 - minor code style fixes
 
 ## [0.2.1] - 2020-12-07
+
 ### Changed
+
 - Metrics are now logged into CSV files: `train.csv`, `val.csv`
 - xlabels of metrics plots in epochs
 - fixed cropped xlabel in precision recall plot
 
 ## [0.2.0] - 2020-11-19
+
 ### Changed
+
 - fixed wrong accuracy metrics
 
 ## [0.1.1] - 2020-10-30
+
 ### Added
+
 - Slope visualization in tile-predictions
 
 ### Changed
+
 - Color scheme in tile-predictions
 - fixed printout of "label" to "skipped" during preprocessing stage of inference  
 

README.md

@@ -2,12 +2,33 @@
 
 ## Installation
 
+### Environment setup
+
+#### Python / conda
+
+We recommend using a new conda environment from scratch
+
+```bash
+conda env create -n thaw_slump_segmentation python=3.10 mamba -c conda-forge
+conda activate thaw_slump_segmentation
+```
+
+#### gdal
+
+gdal incl. gdal-utilities (preferably version >=3.6) need to be installed in your environment, e.g. with conda/mamba
+
+```bash
+mamba install gdal>=3.6 -c conda-forge
+```
+
+### Package Installation
+
 * Latest development version: `pip install git+https://github.com/initze/thaw-slump-segmentation`
 * Latest release: `pip install https://github.com/initze/thaw-slump-segmentation/releases/download/untagged-f6739f56e0ee4c2c64fe/thaw_slump_segmentation-0.10.0-py3-none-any.whl`
 
 This will pull the CUDA 12 version of pytorch. If you are running CUDA 11, you need to manually switch to the corresponding Pytorch package afterwards by running `pip3 install torch==2.2.0+cu118 torchvision==0.17.0+cu118 --index-url https://download.pytorch.org/whl/cu118`
 
-gdal incl. gdal-utilities (preferably version >=3.6) need to be installed in your environment, e.g. with conda
+
 
 ### Additional packages
 #### cucim
@@ -29,9 +50,11 @@ The container contains all requirements to run the processing code, singularity
 singularity pull library://initze/aicore/thaw_slump_segmentation
 singularity shell --nv --bind <your bind path> thaw_slump_segmentation.sif
 ```
+
 ### Option 2 - anaconda
 ### Environment setup
 We recommend using a new conda environment from the provided environment.yml file
+
 ```bash
 conda env create -n aicore -f environment.yml
 ```
@@ -61,27 +84,32 @@ gdal_bin: '%CONDA_PREFIX%\Library\bin' # must be single quote
 ### Data Preprocessing for Planet data
 
 #### Setting up all required files for training and/or inference 
+
 ```bash
 python setup_raw_data.py --data_dir <DATA_DIR>
 ```
 
 #### Setting up required files for training 
+
 ```bash
 python prepare_data.py --data_dir <DATA_DIR>
 ```
 
 ### Data Preprocessing for Sentinel 2 data to match planet format
+
 ```bash
 python download_s2_4band_planet_format.py --s2id <IMAGE_ID> --data_dir <DATA_DIR>
 ```
 
 
 ### Training a model
+
 ```bash
 python train.py --data_dir <DATA_DIR> -n <MODEL_NAME>
 ```
 
 ### Running Inference
+
 ```bash
 python setup_raw_data.py --data_dir <DATA_DIR> --nolabel
 python inference.py --data_dir <DATA_DIR> --model_dir <MODEL_NAME> 20190727_160426_104e 20190709_042959_08_1057

pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "thaw-slump-segmentation"
-version = "0.10.1"
-description = "Add your description here"
+version = "0.10.2"
+description = "Thaw slump segmentation workflow using PlanetScope data and pytorch"
 authors = [
     { name = "Ingmar Nitze", email = "[email protected]" },
     { name = "Konrad Heidler", email = "[email protected]" }