Problem
Research questions and friction points this paper is trying to address.
Autonomous Vehicles
Anomaly Detection
Urban Environment
Self-supervised Likelihood Estimation with Energy Guidance for Anomaly Segmentation in Urban Scenes
🤖 AI Summary
To address the challenges of fine-grained annotation scarcity and absence of anomalous samples in urban-scene anomaly segmentation, this paper proposes an energy-guided self-supervised likelihood estimation framework. Methodologically, it introduces a task-agnostic binary likelihood discriminator and a task-oriented joint energy-residual modeling mechanism, augmented with a context-aware anomaly head and adaptive self-supervised training for dynamic optimization of anomaly masks. Its key contribution lies in the first decoupled modeling of energy functions and pixel-wise likelihood estimation—enabling end-to-end anomaly segmentation without requiring anomalous samples, auxiliary data, or synthetic data. Evaluated on Fishyscapes and Road Anomaly benchmarks, the method achieves performance on par with fully supervised approaches, while significantly enhancing robustness and generalization in zero-shot anomaly detection.
📝 Abstract
Robust autonomous driving requires agents to accurately identify unexpected areas (anomalies) in urban scenes. To this end, some critical issues remain open: how to design advisable metric to measure anomalies, and how to properly generate training samples of anomaly data? Classical effort in anomaly detection usually resorts to pixel-wise uncertainty or sample synthesis, which ignores the contextual information and sometimes requires auxiliary data with fine-grained annotations. On the contrary, in this paper, we exploit the strong context-dependent nature of segmentation task and design an energy-guided self-supervised frameworks for anomaly segmentation, which optimizes an anomaly head by maximizing likelihood of self-generated anomaly pixels. For this purpose, we design two estimators to model anomaly likelihood, one is a task-agnostic binary estimator and the other depicts the likelihood as residual of task-oriented joint energy. Based on proposed estimators, we devise an adaptive self-supervised training framework, which exploits the contextual reliance and estimated likelihood to refine mask annotations in anomaly areas. We conduct extensive experiments on challenging Fishyscapes and Road Anomaly benchmarks, demonstrating that without any auxiliary data or synthetic models, our method can still achieves comparable performance to supervised competitors. Code is available at https://github.com/yuanpengtu/SLEEG.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Energy-guided Self-supervised Estimation
Adaptive Learning Framework
Anomaly Detection in Urban Environments
🔎 Similar Papers
No similar papers found.
