TLS Treelist Prediction

Predict tree plant functional type, genus, and species from terrestrial lidar scans tree features using hyper-param tuned CatBoost trained on FastFuel dataset

Getting started

1. Download the data and form the directory structure shown as follows.

sprint_4/
├── FastFuel
│   ├── FF_treelist_all.csv
│   ├── blk_plot_identification.csv
│   └── fftl_column_description.csv
├── SPCD_from_points.ipynb
├── TLS_catboost.ipynb
├── fia-database-california
│   ├── CA_PLOT.csv
│   └── CA_TREE.csv
├── field_data
│   ├── 01_plot_identification.csv
│   └── 03_tree.csv
├── requirements.txt
├── species_reference
│   ├── FIATreeSpeciesCode_pft.csv
│   └── REF_SPECIES.csv
├── terrestrial-lidar-scans-tls-and-derived-tree-lists-for-field-sampled-plots-for-uc-climate-actio
│   ├── TLS_treelist.csv
│   ├── blk_plot_identification.csv
│   ├── intellimon_chm.zip
│   ├── intellimon_column_descriptions.csv
│   └── tls_files_download_paths.txt
└── tls_catboost_v3.cbm

2. Create environment & install:

   pip install -r requirements.txt

3. Run the notebook

   Please Note:

   • The notebook currently loads trained model weights.

   • To replicate the training process, set is_infer = False in the CFG class.

   • To perform hyperparameter tuning, set n_hyper_trial > 0.

   • Additional details and explanations are included as comments within the notebook.

Results

Acknowledgements

Special thanks to the organizers for organizing the event, providing the dataset and framework for this task.