Co-Seg: An Image Segmentation Framework Against Label Corruption

TL;DR

Co-Seg is a novel framework that enhances image segmentation robustness against label noise by collaboratively training networks, filtering reliable samples, and correcting labels to maintain high performance.

Contribution

The paper introduces Co-Seg, a new collaborative training framework that effectively handles noisy labels in image segmentation tasks, improving robustness without requiring noise-free data.

Findings

Achieves comparable results to noise-free supervised training.

Effectively filters and corrects noisy labels during training.

Easily integrated into existing segmentation algorithms.

Abstract

Supervised deep learning performance is heavily tied to the availability of high-quality labels for training. Neural networks can gradually overfit corrupted labels if directly trained on noisy datasets, leading to severe performance degradation at test time. In this paper, we propose a novel deep learning framework, namely Co-Seg, to collaboratively train segmentation networks on datasets which include low-quality noisy labels. Our approach first trains two networks simultaneously to sift through all samples and obtain a subset with reliable labels. Then, an efficient yet easily-implemented label correction strategy is applied to enrich the reliable subset. Finally, using the updated dataset, we retrain the segmentation network to finalize its parameters. Experiments in two noisy labels scenarios demonstrate that our proposed model can achieve results comparable to those obtained from supervised learning trained on the noise-free labels. In addition, our framework can be easily implemented in any segmentation algorithm to increase its robustness to noisy labels.