Rotation-Robust Regression with Convolutional Model Trees

TL;DR

This paper introduces rotation-robust learning methods for image inputs using Convolutional Model Trees, incorporating geometry-aware biases and deployment-time orientation search to improve robustness against in-plane rotations.

Contribution

It proposes three geometry-aware inductive biases for split directions in CMTs and evaluates a deployment-time orientation search, advancing rotation robustness in image regression tasks.

Findings

Orientation search enhances robustness under severe rotations.

Geometry-aware biases improve in-plane rotation robustness.

Confidence-based orientation selection has both benefits and limitations.

Abstract

We study rotation-robust learning for image inputs using Convolutional Model Trees (CMTs) [1], whose split and leaf coefficients can be structured on the image grid and transformed geometrically at deployment time. In a controlled MNIST setting with a rotation-invariant regression target, we introduce three geometry-aware inductive biases for split directions -- convolutional smoothing, a tilt dominance constraint, and importance-based pruning -- and quantify their impact on robustness under in-plane rotations. We further evaluate a deployment-time orientation search that selects a discrete rotation maximizing a forest-level confidence proxy without updating model parameters. Orientation search improves robustness under severe rotations but can be harmful near the canonical orientation when confidence is misaligned with correctness. Finally, we observe consistent trends on MNIST digit recognition implemented as one-vs-rest regression, highlighting both the promise and limitations of confidence-based orientation selection for model-tree ensembles.