Robust Activation Map Rectification for Weakly Supervised Volumetric Segmentation: Temporal Coherence as a Free Lunch

2026-07-22

AI Summary

In weakly supervised volumetric segmentation, class activation maps (CAMs) often suffer from inaccurate localization due to noise and catastrophic failures. This work proposes a training-free, prototype-free correction framework that, for the first time, leverages temporal and structural consistency in volumetric data as a free inductive bias. By introducing Variance-Reduced Activation Aggregation (VRAA) and Bidirectional Endpoint Refinement (BER), the method substantially enhances CAM reliability. It integrates high-dimensional random vector modeling with a model-agnostic ensemble strategy, achieving significant performance gains across multiple public benchmarks: up to a 20% increase in Dice score, a 40% improvement in mIoU, and over fivefold acceleration in inference speed.

Abstract

Weakly supervised segmentation relies heavily on class activation maps (CAMs) to initially localize target regions. However, CAMs are often noisy and prone to catastrophic failures. Existing remedies typically introduce additional training stages or prototype learning, increasing computational cost and reducing robustness. In this paper, we propose a training-free prototype-free framework that rectifies unreliable CAMs by exploiting temporal and structural coherence in volumetric data as a free lunch. Our approach is built on two key components. First, we introduce Variance-Reduced Activation Aggregation (VRAA) which suppresses noise and amplify coherent semantic signals. We provide a theoretical justification by modeling CAMs as high-dimensional random vectors and show that aggregation yields provable variance reduction. Second, we design a Bidirectional Extremity Rectification (BER) mechanism that detects and rectifies implausible activations through bidirectional extremity checks, effectively mitigating extreme-value failures without learning additional parameters. Our method is model-agnostic and can be seamlessly integrated with existing pipelines. Extensive experiments on multiple public benchmarks demonstrate substantial improvements over state-of-the-art weakly supervised methods, achieving up to 20% Dice and 40% mIoU gains while reducing inference time by more than 5 times. These results indicate that leveraging coherence as an implicit inductive bias yields a principled and efficient approach to stabilizing weakly supervised volumetric segmentation. Our code will be available.

Problem

• Weakly Supervised Segmentation
• Class Activation Maps
• Volumetric Segmentation
• Temporal Coherence
• Noisy CAMs

Innovation

• Class Activation Map Rectification
• Temporal Coherence
• Weakly Supervised Segmentation
• Variance-Reduced Aggregation
• Model-Agnostic Framework