CVLGOct 28, 2020

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

arXiv:2010.14793v16 citationsHas Code
Originality Incremental advance
AI Analysis

This addresses the challenge of segmentation in scenarios with incorrect or no class labels, offering a robust solution for tasks like salient object detection and general segmentation, though it is incremental as it builds on existing architectures like ResNet101 and DeepLab-v3.

The paper tackles the problem of segmentation without predefined class labels by introducing a class-agnostic segmentation (CAS) loss that clusters regions with similar appearance in a weakly-supervised manner, resulting in performance gains such as a 50% improvement over state-of-the-art methods on salient object detection datasets with low-fidelity data and outperforming cross-entropy loss on general segmentation datasets.

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.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes