Class-Agnostic Segmentation Loss and Its Application to Salient Object Detection and Segmentation

TL;DR

This paper introduces a class-agnostic segmentation loss that learns region clusters without predefined labels, improving salient object detection especially with noisy data, and demonstrating broad applicability across segmentation tasks.

Contribution

The paper proposes a novel class-agnostic segmentation loss that is robust to class imbalance and weak supervision, enhancing segmentation performance across various datasets.

Findings

Outperforms state-of-the-art in low-fidelity training scenarios by ~50%.

Matches or exceeds performance of existing methods with high-fidelity data.

Excels in general segmentation tasks, surpassing cross-entropy loss significantly.

Abstract

In this paper we present a novel loss function, called class-agnostic segmentation (CAS) loss. With CAS loss the class descriptors are learned during training of the network. We don't require to define the label of a class a-priori, rather the CAS loss clusters regions with similar appearance together in a weakly-supervised manner. Furthermore, we show that the CAS loss function is sparse, bounded, and robust to class-imbalance. We apply our CAS loss function with fully-convolutional ResNet101 and DeepLab-v3 architectures to the binary segmentation problem of salient object detection. We investigate the performance against the state-of-the-art methods in two settings of low and high-fidelity training data on seven salient object detection datasets. For low-fidelity training data (incorrect class label) class-agnostic segmentation loss outperforms the state-of-the-art methods on salient object detection datasets by staggering margins of around 50%. For high-fidelity training data (correct class labels) class-agnostic segmentation models perform as good as the state-of-the-art approaches while beating the state-of-the-art methods on most datasets. In order to show the utility of the loss function across different domains we also test on general segmentation dataset, where class-agnostic segmentation loss outperforms cross-entropy based loss by huge margins on both region and edge metrics.