Noise-Robust Tiny Object Localization with Flows
This addresses a persistent performance gap in object detection for tiny objects, which is a domain-specific problem in computer vision.
The paper tackles the problem of tiny object localization being highly sensitive to annotation noise, which causes overfitting when optimizing strict localization objectives. The proposed TOLF framework uses normalizing flows for error modeling and uncertainty-guided optimization, achieving a 1.2% AP boost over the DINO baseline on the AI-TOD dataset.
Despite significant advances in generic object detection, a persistent performance gap remains for tiny objects compared to normal-scale objects. We demonstrate that tiny objects are highly sensitive to annotation noise, where optimizing strict localization objectives risks noise overfitting. To address this, we propose Tiny Object Localization with Flows (TOLF), a noise-robust localization framework leveraging normalizing flows for flexible error modeling and uncertainty-guided optimization. Our method captures complex, non-Gaussian prediction distributions through flow-based error modeling, enabling robust learning under noisy supervision. An uncertainty-aware gradient modulation mechanism further suppresses learning from high-uncertainty, noise-prone samples, mitigating overfitting while stabilizing training. Extensive experiments across three datasets validate our approach's effectiveness. Especially, TOLF boosts the DINO baseline by 1.2% AP on the AI-TOD dataset.