Deep Learning and Energy Models for Fine Dead Wood Segmentation - Jacquelyn Shelton

Explore deep learning and energy models for fine dead wood segmentation in this conference talk from the Machine Learning for Climate KITP conference. Delve into the carbon cycle, the importance of dead trees, and the study area data used for basic segmentation tasks. Learn about various approaches including unit regression, centroids, and Mask RCNN. Examine the multi-term energy model, incorporating image and shape terms for multiple contours. Analyze experimental results, comparing recall and precision metrics. Gain insights into future work in this field and participate in a Q&A session to further understand the application of machine learning in climate science.

Introduction
The Carbon Cycle
Why Dead Trees
Data
Study area
Data used
Basic segmentation tasks
Approach
Unit
Regression
Centroids
Mask RCNN
Multiterm Energy Model
Image Term
Shape Term
Multiple Contours
Experiment
Training polygons
Metrics
Results
Recall and precision
Comparison
Summary
Future work
Thank you
Space Invaders
Masks
Pipelines
Questions