With the rapid development of tunnel and underground engineering projects in China, the safety assessment and routine maintenance of lining structures have become increasingly critical. Common defects such as cracks, segment damage, secondary lining spalling, and water leakage, if not detected and repaired in a timely manner, can seriously threaten the durability and operational safety of tunnels. However, traditional manual inspection methods suffer from low efficiency and strong subjectivity, underscoring the urgent need to leverage intelligent approaches to improve detection accuracy and efficiency.

To advance the intelligent safety assessment of tunnel and underground structures, on the occasion of the 10th anniversary of the Underground Space journal and during the 3rd Jeme Tien Yow Lecture, we have launched a student contest: "Intelligent Multi-defect Detection for Tunnel Linings in Complex Environments." The winning teams' original papers from this contest will be invited to submit to Underground Space.

To address the inefficiency and subjectivity inherent in traditional manual inspections, and to overcome the challenges associated with deploying existing deep learning models in complex tunnel environments and on mobile platforms, this contest sets forth the following:

•  
  Core Task: Participants are required to develop a lightweight, high-precision semantic segmentation model based on the provided dataset of real-world tunnel lining images.
•  
  Detection Targets: The model must achieve pixel-level classification for four typical types of defects: cracks, segment damage, secondary lining spalling, and water leakage.
•  
  Performance Requirements: While ensuring high detection accuracy, the model must exhibit rapid inference speed and a compact size to meet the deployment demands of mobile or embedded devices.

Date: June 6-7, 2026

Venue: Tongji University, 1239 Siping Road, Yangpu District, Shanghai

Registration Fee: Free of charge

Organizer:
Chinese Society of Rock Mechanics and Engineering
Underground Space Journal

Co-organizer:
- Chinese Institution of Information Technology and Application in Geo-Engineering,
  Chinese Society of Rock Mechanics and Engineering
- Tongji University
- Engineering Research Center of Civil Informatics, Ministry of Education, Tongji University

Honorary Chairs:
- Hehua Zhu      Antonio Bobet      Teruo Nakai

Chairs:
- Feng Zhang      Lianyang Zhang      Xiaojun Li

Vice Chairs:
- Chenjie Gong      Mengqi Zhu

•  
  Dataset Scale: The organizing committee will provide a dataset consisting of images with detailed annotations.
•  
  Data Characteristics: The dataset authentically reflects complex field conditions, featuring uneven illumination, significant background interference, and diverse surface textures. The images are provided in JPG and PNG formats. The annotations are in JSON format generated via LabelMe.
•  
  On-site Testing: The final evaluation will be conducted on-site utilizing a private test set. Participants are required to bring their trained models, generate inference results at the competition venue, and submit them immediately.

Participating teams are required to submit the following four deliverables:

•  
  Complete Code Package: Includes data preprocessing, model definition, and inference scripts.
•  
  Trained Model Weights: .pth, .onnx, .pb, or other standard formats.
•  
  Technical Report: Please submit a technical report in both PDF and Word formats. The content should cover model architecture, training strategy, and experimental results and analysis based on the provided 1,100 annotated images. All content must be written in English.
•  
  Inference Results (Submitted On-site): Instead of an execution script, participants are required to submit the inference result files generated during the on-site test. Participants must bring their trained models and generate the inference results on the designated test set provided at the competition venue. JSON files contain predicted shape information and class labels. The generated JSON files must be submitted immediately after the inference process.

The final score is comprehensively determined by detection accuracy, model size, and expert evaluation.

•  
  Formula: Score = 0.5×Accuracy Score × Size Coefficient + 0.5 × ExpertScore.
•  
  Accuracy Score: mIoU across 7 classes on the test set × 100.
•  
  Size Coefficient: 1.1 for models ≤50MB; 1.0 for 50MB-100MB; 0.9 for models ≥100MB.
•  
  ExpertScore: Calculated as the arithmetic mean of scores awarded by review experts based on the technical documentation and on-site defense.