Visual Accommodation: Rethinking Image Scale as a Learnable Variable for Object Detection

arXiv:2412.06341v2 Announce Type: replace-cross Abstract: We propose Ciliary-DETR (previous name: Elastic-DETR), a framework for test-time resolution adjustment analogous to biological accommodation. While multi-scale data augmentation improves robustness to scale variation, modern detectors rely on fixed inference resolutions, potentially limiting flexibility and robustness. Similar to the ciliary muscle, we introduce a lightweight scale predictor that dynamically estimates test-time scale fact