Beyond the Pixel-Wise Loss for Topology-Aware Delineation
This work addresses the challenge of accurate delineation for applications like microscopy and aerial imaging, offering a novel approach that improves upon existing methods, though it is incremental in its refinement of loss functions.
The paper tackles the problem of delineating curvilinear structures in computer vision by addressing the limitations of pixel-wise losses, which fail to account for topological errors, and introduces a new topology-aware loss term and iterative refinement pipeline that significantly boost accuracy, nearly doubling it in some cases compared to standard methods.
Delineation of curvilinear structures is an important problem in Computer Vision with multiple practical applications. With the advent of Deep Learning, many current approaches on automatic delineation have focused on finding more powerful deep architectures, but have continued using the habitual pixel-wise losses such as binary cross-entropy. In this paper we claim that pixel-wise losses alone are unsuitable for this problem because of their inability to reflect the topological impact of mistakes in the final prediction. We propose a new loss term that is aware of the higher-order topological features of linear structures. We also introduce a refinement pipeline that iteratively applies the same model over the previous delineation to refine the predictions at each step while keeping the number of parameters and the complexity of the model constant. When combined with the standard pixel-wise loss, both our new loss term and our iterative refinement boost the quality of the predicted delineations, in some cases almost doubling the accuracy as compared to the same classifier trained with the binary cross-entropy alone. We show that our approach outperforms state-of-the-art methods on a wide range of data, from microscopy to aerial images.