AI FOR CULTURAL HERITAGE
Research on Batik Image Pattern Detection Based on Improved YOLOv11
Batik, a crucial intangible cultural heritage, features intricate pattern systems. However, detecting these patterns in complex batik images is challenging due to dense distributions, scale variations, and degraded quality. This paper introduces a robust batik pattern detection model, an improved YOLOv11 architecture that balances high detection accuracy with computational efficiency.
Executive Impact: Enhanced Cultural Heritage AI
Our improved YOLOv11 model offers a cutting-edge solution for the digital preservation and interpretation of valuable batik heritage, overcoming previous limitations with significant performance gains.
4.47%
mAP Improvement (Ablation)
49.6%
Max mAP Improvement (Comparative)
370
Frames Per Second (FPS)
9,933
Annotated Bounding Boxes
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Enhanced YOLOv11 Architecture
Our model significantly improves the YOLOv11 base through two key enhancements: Vision Outlooker (VOLO) attention and Fused-MBConv modules. VOLO addresses the limitations of convolutional neural networks in capturing long-distance spatial dependencies crucial for complex batik patterns, enhancing global perception while maintaining computational efficiency. The Fused-MBConv modules optimize feature extraction by synergistically combining depth-wise convolutions, inverted residual structures, and SE Layers within the C3K2 architecture. This reduces network complexity while boosting feature quality and robustness against variations in real batik samples.
Comprehensive Dataset & Robust Evaluation
To overcome data scarcity, we constructed the largest and most comprehensive Chinese batik pattern dataset, comprising 861 images with 9933 meticulously annotated bounding boxes across 7 categories. Extensive data augmentation techniques (Blur, CLAHE, Flip, Mosaic, etc.) were applied to enhance model generalizability and prevent overfitting. Ablation studies demonstrated a 4.47% mAP improvement from integrating VOLO and Fused-MBConv. Comparative analyses against lightweight models like YOLOv8n, YOLOv9t, YOLOv10n, PP-YOLOE-Lite, Ghost YOLO, and MobileVit YOLO confirmed our model's superior detection accuracy (up to 49.6% higher mAP) while maintaining efficient inference speeds suitable for diverse applications.
Integrated Batik Detection and Analysis System
We developed a prototype system that seamlessly integrates our improved YOLOv11 model with a batik knowledge graph. This system offers end-to-end functionality from visual recognition to semantic analysis. It features an image detection module for uploading batik images, automatic pattern detection, and visualization with category annotations and statistical analysis. The knowledge graph module enables deep semantic exploration, providing users with pattern-related nodes, similar examples, and cultural information including prototype sources, symbolic meanings, and worship consciousness. This holistic approach supports systematic documentation, cultural education, and restoration efforts for intangible cultural heritage.
4.47%
Increase in Mean Average Precision (mAP) through architectural enhancements.
Enterprise Process Flow
Chinese Batik Dataset Construction & Augmentation
→
YOLOv11 Baseline Model
→
VOLO Attention Integration (C2PSA)
→
Fused-MBConv Optimization (C3K2)
→
Improved YOLOv11 Model
→
Integrated Detection & Analysis System
| Indicator | MobileNetv2-SSD | YOLOv8n | YOLOv9t | YOLOv10n | PP-YOLOE-Lite | Ghost YOLO | MobileVit YOLO | Ours |
|---|---|---|---|---|---|---|---|---|
| P | 0.773 | 0.747 | 0.731 | 0.701 | 0.784 | 0.688 | 0.536 | 0.733 |
| R | 0.198 | 0.679 | 0.669 | 0.639 | 0.671 | 0.610 | 0.581 | 0.699 |
| mAP | 0.500 | 0.726 | 0.720 | 0.680 | 0.720 | 0.661 | 0.584 | 0.748 |
| FPS/f.s-1 | 146 | 453 | 286 | 450 | 247 | 476 | 157 | 370 |
| Parameters/million | 3.926 | 2.686 | 1.731 | 2.700 | 8.364 | 1.394 | 8.580 | 3.379 |
| Model size/MB | 18.283 | 5.526 | 4.100 | 5.655 | 16.788 | 3.083 | 17.097 | 6.962 |
Our model demonstrates superior overall performance, balancing high detection accuracy (achieving the highest mAP) with efficient inference speeds, positioning it as the most effective solution for complex batik pattern detection among lightweight models.
Case Study: Digital Preservation of Miao Batik
The developed Batik Detection and Analysis System provides a comprehensive platform for the digital preservation and interpretation of intangible cultural heritage. This system allows cultural institutions to upload complex batik images, automatically detect intricate patterns, and access rich cultural knowledge. By linking visual detection with a detailed knowledge graph, the system reveals symbolic meanings, historical contexts, and associated worship consciousness of the detected patterns.
For example, if a "dragon" pattern is detected, the system immediately provides insights into its significance in Miao culture, its prototype sources, and related cultural taboos. This automated semantic analysis capability vastly reduces manual annotation efforts, streamlines archival processes, and enforces public engagement for exhibitions and educational programs, ensuring the sustainable inheritance of this invaluable heritage.
Calculate Your Potential AI Impact
Estimate the efficiency gains and cost savings your enterprise could achieve by integrating our advanced AI solutions.
Your AI Implementation Roadmap
A structured approach to integrating advanced AI into your enterprise, ensuring seamless adoption and measurable success.
Phase 01: Discovery & Strategy
In-depth analysis of current workflows, identification of AI opportunities, and development of a tailored implementation strategy aligning with your business objectives.
Phase 02: Solution Design & Prototyping
Designing the AI architecture, selecting optimal models (like improved YOLOv11), and developing initial prototypes to validate core functionalities and performance.
Phase 03: Development & Integration
Full-scale development of the AI solution, seamless integration into existing enterprise systems, and rigorous testing for performance, security, and scalability.
Phase 04: Deployment & Optimization
Go-live support, continuous monitoring of AI model performance, and iterative optimization based on real-world data and feedback to maximize impact.
Ready to Transform Your Enterprise with AI?
Our experts are ready to guide you through integrating cutting-edge AI solutions. Book a complimentary consultation to discuss your specific needs and strategic goals.
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