A new architecture-diverse TabNet ensemble classifies seven crop types from harmonized multi-source soil data with 90.5 percent accuracy, outperforming XGBoost and single optimized models.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results