Revisiting Generative Adversarial Networks for Binary Semantic Segmentation on Imbalanced Datasets

TL;DR

This work tackles binary semantic segmentation of anomalous pavement cracks in highly imbalanced datasets. It introduces a two-stage conditional GAN framework with a UNet-like generator, dual discriminators, and an auxiliary network, augmented by attention gates and entropy-based losses to produce robust, multiscale crack probability maps. The approach yields state-of-the-art results across six pavement datasets (e.g., CRACK500, CrackTree26, CrackLS315, CFD, CRKWH100, DeepCrack-DB) while maintaining practical computation; ablations show benefits from CBAM/LSA attention and Tversky-type losses. Overall, the method enhances crack detection under diverse conditions, supporting scalable, automatic pavement maintenance.

Abstract

Anomalous crack region detection is a typical binary semantic segmentation task, which aims to detect pixels representing cracks on pavement surface images automatically by algorithms. Although existing deep learning-based methods have achieved outcoming results on specific public pavement datasets, the performance would deteriorate dramatically on imbalanced datasets. The input datasets used in such tasks suffer from severely between-class imbalanced problems, hence, it is a core challenge to obtain a robust performance on diverse pavement datasets with generic deep learning models. To address this problem, in this work, we propose a deep learning framework based on conditional Generative Adversarial Networks (cGANs) for the anomalous crack region detection tasks at the pixel level. In particular, the proposed framework containing a cGANs and a novel auxiliary network is developed to enhance and stabilize the generator's performance under two alternative training stages, when estimating a multiscale probability feature map from heterogeneous and imbalanced inputs iteratively. Moreover, several attention mechanisms and entropy strategies are incorporated into the cGANs architecture and the auxiliary network separately to mitigate further the performance deterioration of model training on severely imbalanced datasets. We implement extensive experiments on six accessible pavement datasets. The experimental results from both visual and quantitative evaluation show that the proposed framework can achieve state-of-the-art results on these datasets efficiently and robustly without acceleration of computation complexity.

Figures (13)

• Figure 1: The general schema of our proposed method.
• Figure 2: The generator architecture of our proposed method.
• Figure 3: Two discriminator architectures.
• Figure 4: CBAM attention architectures.
• Figure 5: Local self-attention (LSA) layer architectures.