AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

Contrasting Local and Global Modeling with Machine Learning and Satellite Data: A Case Study Estimating Tree Canopy Height in African Savannas

Esther Rolf, Lucia Gordon, Milind Tambe, Andrew Davies

JMLR 2026 · Learning theory and algorithms › Prediction and decision losses · approach: Provable post-hoc explanations · 8 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
train local FCN on local labels instead of using the best global mapLocal FCN vs Meta/ETH/GLAD maps · Karingani TCHRMSE: 4.51 -> 1.64 (ETH); 3.09 -> 1.65 (GLAD); 3.51 -> 1.65 (Meta); 2.43 -> 1.65Improved
initialize XceptionS2 from global weights instead of random init (ablation)XceptionS2 · Karingani TCHRMSE: 1.98 vs 2.05 (random), 1.94 (random, no latlon)No clear effect
initialize U-Net with SSL4EO self-supervised weights (ablation)U-Net ResNet-18 · Karingani TCHRMSE: 2.27 -> 2.05Improved
fine-tune entire U-Net instead of only the decoder (ablation)U-Net SSL init · Karingani TCHRMSE: 2.05 -> 2.08No clear effect
remove location embedding from XceptionS2 (ablation)XceptionS2 random init · Karingani TCHRMSE: 2.05 -> 1.94Improved
use small FCN instead of globally initialized modelsLocal FCN vs best globally initialized model · Karingani TCHRMSE: 0.34m lower average RMSE, 0.28m lower MAEImproved
add near-infrared band to RGB input (ablation)Local FCN · Karingani TCHRMSE: about 10% lower average RMSEImproved
add remaining 8 Sentinel-2 bands (ablation)Local FCN · Karingani TCHRMSE: small positive effect; 3-channel vs 12-channel gap 0.22mImproved
add 3-channel location embeddings (ablation)Local FCN · Karingani TCHRMSE: did not change performance greatly, if anything decreasedNo clear effect
increase number of training sites from 3 to 12 (ablation)Local FCN · Karingani TCHRMSE: 0.25m average gap between 3 and 12 sitesImproved
use FCN instead of XceptionS2 (ablation)Local models · Karingani TCHRMSE: 0.30m gap in average RMSEImproved
use local FCN predictions instead of ETH/GLAD maps for biomass calibrationAGB estimation · Aboveground biomass calibrationR2: 0.11 (ETH) -> 0.52; 0.22 (GLAD) -> 0.52Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy