Bridging Synthetic and Real-World Domains: A Human-in-the-Loop Weakly-Supervised Framework for Industrial Toxic Emission Segmentation

2025-07-29

AI Summary

Industrial smoke segmentation faces practical deployment challenges due to the high cost of pixel-level annotation and scarcity of real-world labeled data. To address this, we propose a novel framework integrating citizen participation with weakly supervised domain adaptation. First, video-level citizen annotations are leveraged to refine pseudo-labels for synthetic data generation via a multi-vote citizen voting mechanism. Second, we design a class-aware adversarial domain adaptation module, employing class-specific discriminators to align source and target domain features—enabling effective segmentation without any pixel-level labels in the target domain. Evaluated on SMOKE5K and IJmond benchmarks, our method achieves F1 = 0.414 and smoke IoU = 0.261—5–6× higher than baseline weakly supervised methods—and matches the performance of a fully supervised model trained on only 100 densely annotated images. This significantly enhances the practicality and scalability of smoke segmentation in real industrial settings.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Application Domains: Internet of Things, Sensor Networks & Smart Cities

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Industrial smoke segmentation is critical for air-quality monitoring and environmental protection but is often hampered by the high cost and scarcity of pixel-level annotations in real-world settings. We introduce CEDANet, a human-in-the-loop, class-aware domain adaptation framework that uniquely integrates weak, citizen-provided video-level labels with adversarial feature alignment. Specifically, we refine pseudo-labels generated by a source-trained segmentation model using citizen votes, and employ class-specific domain discriminators to transfer rich source-domain representations to the industrial domain. Comprehensive experiments on SMOKE5K and custom IJmond datasets demonstrate that CEDANet achieves an F1-score of 0.414 and a smoke-class IoU of 0.261 with citizen feedback, vastly outperforming the baseline model, which scored 0.083 and 0.043 respectively. This represents a five-fold increase in F1-score and a six-fold increase in smoke-class IoU. Notably, CEDANet with citizen-constrained pseudo-labels achieves performance comparable to the same architecture trained on limited 100 fully annotated images with F1-score of 0.418 and IoU of 0.264, demonstrating its ability to reach small-sampled fully supervised-level accuracy without target-domain annotations. Our research validates the scalability and cost-efficiency of combining citizen science with weakly supervised domain adaptation, offering a practical solution for complex, data-scarce environmental monitoring applications.

Problem

• Industrial smoke segmentation lacks pixel-level annotations
• Weakly-supervised domain adaptation integrates citizen-provided labels
• Achieves high accuracy without target-domain annotations

Innovation

• Human-in-the-loop weakly-supervised domain adaptation
• Class-aware adversarial feature alignment
• Citizen feedback refined pseudo-labels