Fusion Complexity Inversion: Why Simpler Cross View Modules Outperform SSMs and Cross View Attention Transformers for Pasture Biomass Regression
This research provides actionable guidelines for improving pasture biomass estimation from scarce agricultural imagery, which is critical for sustainable livestock management, by demonstrating that simpler fusion modules and stronger backbones are more effective.
This study systematically evaluated vision foundation models for pasture biomass regression on a small, imbalanced dataset, revealing that a simpler two-layer gated depthwise convolution achieved an R^2 of 0.903, outperforming more complex cross-view attention transformers (0.833) and state space models (0.819). The research found that backbone pretraining scale was the dominant factor, with a DINOv2 to DINOv3 upgrade alone yielding a +5.0 R^2 improvement.
Accurate estimation of pasture biomass from agricultural imagery is critical for sustainable livestock management, yet existing methods are limited by the small, imbalanced, and sparsely annotated datasets typical of real world monitoring. In this study, adaptation of vision foundation models to agricultural regression is systematically evaluated on the CSIRO Pasture Biomass benchmark, a 357 image dual view dataset with laboratory validated, component wise ground truth for five biomass targets, through 17 configurations spanning four backbones (EfficientNet-B3 to DINOv3-ViT-L), five cross view fusion mechanisms, and a 4x2 metadata factorial. A counterintuitive principle, termed "fusion complexity inversion", is uncovered: on scarce agricultural data, a two layer gated depthwise convolution (R^2 = 0.903) outperforms cross view attention transformers (0.833), bidirectional SSMs (0.819), and full Mamba (0.793, below the no fusion baseline). Backbone pretraining scale is found to monotonically dominate all architectural choices, with the DINOv2 -> DINOv3 upgrade alone yielding +5.0 R^2 points. Training only metadata (species, state, and NDVI) is shown to create a universal ceiling at R^2 ~ 0.829, collapsing an 8.4 point fusion spread to 0.1 points. Actionable guidelines for sparse agricultural benchmarks are established: backbone quality should be prioritized over fusion complexity, local modules preferred over global alternatives, and features unavailable at inference excluded.