fANOVA Quantifies Design Choice Impact in Remote Sensing MLC

This study uses functional analysis of variance (fANOVA) to quantify how architecture, fine-tuning, and initialization impact multi-label classification in remote sensing. Results show that datasets group naturally based on their sensitivity to specific design decisions rather than model rankings alone.