Learning Camouflaged Object Detection from Noisy Pseudo Label

TL;DR

This paper introduces a semi-supervised camouflaged object detection method that effectively learns from noisy pseudo labels generated from limited fully labeled data, achieving high accuracy with less annotation effort.

Contribution

It proposes the first weakly semi-supervised COD approach using box prompts and a noise correction loss to handle noisy pseudo labels from limited data.

Findings

Outperforms state-of-the-art methods with only 20% fully labeled data.

Uses noise correction loss to improve learning from noisy pseudo labels.

Achieves high-precision camouflaged object segmentation with limited annotations.

Abstract

Existing Camouflaged Object Detection (COD) methods rely heavily on large-scale pixel-annotated training sets, which are both time-consuming and labor-intensive. Although weakly supervised methods offer higher annotation efficiency, their performance is far behind due to the unclear visual demarcations between foreground and background in camouflaged images. In this paper, we explore the potential of using boxes as prompts in camouflaged scenes and introduce the first weakly semi-supervised COD method, aiming for budget-efficient and high-precision camouflaged object segmentation with an extremely limited number of fully labeled images. Critically, learning from such limited set inevitably generates pseudo labels with serious noisy pixels. To address this, we propose a noise correction loss that facilitates the model's learning of correct pixels in the early learning stage, and corrects the error risk gradients dominated by noisy pixels in the memorization stage, ultimately achieving accurate segmentation of camouflaged objects from noisy labels. When using only 20% of fully labeled data, our method shows superior performance over the state-of-the-art methods.