Deep Learning with Airborne Laser Scanning Data for Forest Inventories

This tutorial provides an introductory, hands-on guide to applying deep learning techniques to Enhanced Forest Inventories (EFIs) using airborne laser scanning (ALS) data. Developed for the Canadian Cross-Country EFI Checkup, the workshop covers the complete workflow: from understanding deep learning concepts, preparing and reading data, to training, evaluating, and deploying deep learning models for forest inventory tasks such as species classification and biomass prediction.

Key Features

Intro to Deep Learning for EFIs: Learn the fundamentals of deep learning and its applications in forest inventory.

Data Preparation: Step-by-step instructions for preparing ALS data for deep learning workflows.

PyTorch Lightning: Practical examples using the PyTorch Lightning library for efficient model development.

Model Training & Evaluation: Guided exercises on training, testing, and evaluating deep learning models.

Real-World Applications: Case studies demonstrating deep learning solutions for forest inventory challenges.

Data & Resources

All datasets, training outputs, and model evaluations used in this tutorial are precomputed and provided for download to ensure a smooth learning experience. This allows participants to focus on the workflow without long training times.

Workshop Codebase: GitHub Repository

Tutorial Website: Workshop Site

Original Dataset

This tutorial uses a modified version of the Petawawa Research Forest (PRF) dataset hosted by the Canadian National Forest Information System. This dataset can be found here. For more information about the PRF dataset, see White et al.  (2019) below.

White, Joanne C., et al. "The Petawawa Research Forest: Establishment of a remote sensing supersite." The Forestry Chronicle 95.3 (2019): 149-156.