A Likelihood Ratio-Based Approach to Segmenting Unknown Objects

TL;DR

This paper introduces a lightweight likelihood-ratio-based method using an adaptive unknown estimation module to improve out-of-distribution segmentation of unknown objects, achieving state-of-the-art results without retraining the core model.

Contribution

It proposes a novel, lightweight unknown estimation module and a likelihood-ratio scoring function that enhance OoD segmentation without disrupting the foundational model's learned features.

Findings

Achieves 5.74% higher average precision than previous methods.

Maintains strong inlier segmentation performance.

Reduces false-positive rate in OoD detection.

Abstract

Addressing the Out-of-Distribution (OoD) segmentation task is a prerequisite for perception systems operating in an open-world environment. Large foundational models are frequently used in downstream tasks, however, their potential for OoD remains mostly unexplored. We seek to leverage a large foundational model to achieve robust representation. Outlier supervision is a widely used strategy for improving OoD detection of the existing segmentation networks. However, current approaches for outlier supervision involve retraining parts of the original network, which is typically disruptive to the model's learned feature representation. Furthermore, retraining becomes infeasible in the case of large foundational models. Our goal is to retrain for outlier segmentation without compromising the strong representation space of the foundational model. To this end, we propose an adaptive, lightweight unknown estimation module (UEM) for outlier supervision that significantly enhances the OoD segmentation performance without affecting the learned feature representation of the original network. UEM learns a distribution for outliers and a generic distribution for known classes. Using the learned distributions, we propose a likelihood-ratio-based outlier scoring function that fuses the confidence of UEM with that of the pixel-wise segmentation inlier network to detect unknown objects. We also propose an objective to optimize this score directly. Our approach achieves a new state-of-the-art across multiple datasets, outperforming the previous best method by 5.74% average precision points while having a lower false-positive rate. Importantly, strong inlier performance remains unaffected.