Enterprise AI Analysis
Automated defect classification and localization in sewer pipelines using hybrid ResNet50-Swin transformer and modified YOLOv8 on CCTV inspection images
This study presents a robust two-stage deep learning framework for automated defect classification and localization in sewer pipelines. It uses a hybrid ResNet50-Swin Transformer for initial image classification, achieving 90.28% accuracy, which significantly reduces false positives for the subsequent detection stage. For localization, a modified YOLOv8 integrated with Convolutional Block Attention Modules (CBAM) achieves an 11% mAP gain to 0.81, particularly effective for subtle defects. Validated on 6,912 images from Iranian sewer systems, this scalable, real-time solution outperforms baseline methods in efficiency and accuracy, aligning with urban infrastructure monitoring needs.
Executive Impact
Key performance indicators demonstrating the immediate value of this AI solution.
0
Classification Accuracy
0
Localization Precision
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Problem Statement
Core Innovation
AI Model
Addressing Critical Challenges
Manual evaluation of CCTV footage is inefficient, error-prone, and struggles with low-contrast, noisy, and sediment-heavy images, leading to poor generalization of existing deep learning models.
Cutting-Edge Solution
We present a two-stage approach that first leverages a hybrid ResNet50-Swin Transformer classifier to robustly distinguish defective from non-defective images with 90.28% accuracy, dramatically reducing misclassification and data volume for the subsequent stage. This filtering mechanism directly enhances detection precision-improving the mAP of our modified YOLOv8 from 0.70 to 0.81 (a 11% gain) by eliminating false-positive inputs. For defect localization, we integrate Convolutional Block Attention Modules (CBAM) into YOLOv8, enhancing the model's ability to focus on regions where defect boundaries are difficult to distinguish.
AI Model Architecture: Hybrid ResNet50-Swin Transformer, Modified YOLOv8
The solution combines a Hybrid ResNet50-Swin Transformer for classification and a Modified YOLOv8 with CBAM for precise localization. This architecture is designed for robust performance on low-quality CCTV images, essential for real-world applications in urban infrastructure monitoring.
Enterprise Process Flow
CCTV Video Input
→
Hybrid ResNet50-Swin Classifier (90.28% Acc.)
→
Filter Non-Defective Images
→
Modified YOLOv8 Detector with CBAM (0.81 mAP)
→
Localized Defect Output
Performance vs. Baselines (Classification)
| Model | Accuracy |
|---|---|
| Hybrid ResNet50-Swin Transformer (Ours) | 90.28% |
| ResNet50 | 67.89% |
| Swin Transformer | 75.75% |
| MobileNetV2 | 58.45% |
| InceptionV3 | 68.0% |
| VGG16 | 79.0% |
Performance vs. Baselines (Detection)
| Model | mAP@50 | FPS | Params(M) |
|---|---|---|---|
| Modified YOLOv8n + CBAM (Ours) | 0.81 | 210 | 4.5 |
| YOLOv8n (Baseline) | 0.76 | 230 | 3.2 |
| YOLOv9c | 0.86 | 160 | 25.5 |
| YOLOv10n | 0.74 | 240 | 2.3 |
Real-World Impact in Iran
The system was validated on 6,912 images from over 200 sewer pipelines in Iran, collected under challenging real-world conditions (erosive, humid, poorly illuminated).
Focus on root intrusions, deposits, and open joints, which constitute approximately 87% of all structural issues in polyethylene sewer systems in Iran.
The framework offers a scalable, real-time solution for urban infrastructure monitoring, outperforming baseline methods in efficiency and accuracy, directly supporting Iran's Water and Wastewater Company's maintenance planning.
Advanced ROI Calculator
Estimate your potential cost savings and efficiency gains with an AI-powered solution.
Your AI Implementation Roadmap
A typical journey to integrate advanced AI solutions into your enterprise.
Phase 1: Discovery & Strategy
Initial consultations to understand your unique business needs, data landscape, and strategic objectives. We identify key opportunities for AI integration and define success metrics.
Phase 2: Data Preparation & Model Development
Cleaning, labeling, and augmenting your proprietary datasets. Our engineers then develop, fine-tune, and rigorously test custom AI models tailored to your specific challenges.
Phase 3: Integration & Deployment
Seamlessly integrate the AI solution into your existing infrastructure and workflows. This includes API development, system testing, and user training to ensure smooth adoption.
Phase 4: Monitoring, Optimization & Scaling
Continuous monitoring of model performance, A/B testing, and iterative optimization. We provide ongoing support and strategic guidance for scaling the solution across your enterprise.
Ready to Transform Your Operations with AI?
Schedule a complimentary strategy session with our AI experts to explore how these insights can drive your enterprise forward.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat