Auto Seg-Loss: Searching Metric Surrogates for Semantic Segmentation

TL;DR

This paper introduces Auto Seg-Loss, a method to automatically search for differentiable surrogate loss functions tailored to specific semantic segmentation metrics, improving performance over manual losses.

Contribution

It proposes a novel automated search framework for metric-specific surrogate losses in semantic segmentation, replacing non-differentiable metric components with parameterized functions.

Findings

Searched surrogate losses outperform manual losses on PASCAL VOC and Cityscapes.

The learned losses generalize well across different datasets and network architectures.

The method achieves consistent performance improvements in semantic segmentation tasks.

Abstract

Designing proper loss functions is essential in training deep networks. Especially in the field of semantic segmentation, various evaluation metrics have been proposed for diverse scenarios. Despite the success of the widely adopted cross-entropy loss and its variants, the mis-alignment between the loss functions and evaluation metrics degrades the network performance. Meanwhile, manually designing loss functions for each specific metric requires expertise and significant manpower. In this paper, we propose to automate the design of metric-specific loss functions by searching differentiable surrogate losses for each metric. We substitute the non-differentiable operations in the metrics with parameterized functions, and conduct parameter search to optimize the shape of loss surfaces. Two constraints are introduced to regularize the search space and make the search efficient. Extensive experiments on PASCAL VOC and Cityscapes demonstrate that the searched surrogate losses outperform the manually designed loss functions consistently. The searched losses can generalize well to other datasets and networks. Code shall be released.