Pixel is All You Need: Adversarial Trajectory-Ensemble Active Learning for Salient Object Detection

TL;DR

This paper introduces ATAL, an active learning method that enables training saliency models with minimal point annotations, achieving near fully-supervised performance with only ten points per image.

Contribution

The paper presents a novel adversarial trajectory-ensemble active learning approach that effectively reduces annotation effort while maintaining high saliency detection performance.

Findings

Achieves 97-99% performance of fully-supervised models with only ten points per image.

Proposes an adversarial attack for better uncertainty detection in active learning.

Introduces a relationship-aware diversity sampling algorithm to improve sample selection.

Abstract

Although weakly-supervised techniques can reduce the labeling effort, it is unclear whether a saliency model trained with weakly-supervised data (e.g., point annotation) can achieve the equivalent performance of its fully-supervised version. This paper attempts to answer this unexplored question by proving a hypothesis: there is a point-labeled dataset where saliency models trained on it can achieve equivalent performance when trained on the densely annotated dataset. To prove this conjecture, we proposed a novel yet effective adversarial trajectory-ensemble active learning (ATAL). Our contributions are three-fold: 1) Our proposed adversarial attack triggering uncertainty can conquer the overconfidence of existing active learning methods and accurately locate these uncertain pixels. {2)} Our proposed trajectory-ensemble uncertainty estimation method maintains the advantages of the ensemble networks while significantly reducing the computational cost. {3)} Our proposed relationship-aware diversity sampling algorithm can conquer oversampling while boosting performance. Experimental results show that our ATAL can find such a point-labeled dataset, where a saliency model trained on it obtained $97\%$ -- $99\%$ performance of its fully-supervised version with only ten annotated points per image.