Rotation-Robust Regression with Convolutional Model Trees
This work addresses rotation robustness in image-based regression for applications like digit recognition, but it is incremental as it builds on existing Convolutional Model Trees with specific modifications.
The paper tackled the problem of making regression models robust to image rotations using Convolutional Model Trees, introducing geometry-aware inductive biases and a deployment-time orientation search, which improved robustness under severe rotations but sometimes harmed performance near canonical orientations.
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.