TensionTrackerV1

Estimate wire tension via SAM2-based ROI, CoTracker2-based strain estimation, and edge-based lay angle

System Requirements

All experiments were conducted under the following hardware and software configuration:

Hardware Environment

• Workstation: Dell Precision 7920 Rack
• CPU: Intel Xeon Silver 4210R (10 cores, 20 threads, 2.4–3.2 GHz)
• Memory: 128 GB DDR4-3200 ECC
• GPU: NVIDIA RTX A6000 (48 GB GDDR6 ECC)
• Storage: 1 TB NVMe SSD
• Operating System: Ubuntu 20.04 LTS

Software Environment

• Python: 3.10.13
• CUDA : 12.5

Preparation

• Clone or download the code package from this repository.
• Set the working directory:

MainFolder = "/your/custom/path"

📥 Download Pre-trained Models and Fine Tuning

The following checkpoints are required. Please download and place each file in the specified directory.

🔹 SAM2 (Segment Anything Model V2)

• Checkpoint: sam2.1_hiera_large.pt
• Destination: sam2_repo/checkpoints/
• Download: https://github.com/facebookresearch/sam2
• Config Base: sam2.1_hiera_l.yaml

• multimask_output_in_sam = False
  → Since only a single-object segmentation is required, multi-mask output is disabled.

• num_maskmem = 3
  → Reduced from the default (7) to improve memory efficiency, as the strand shape does not change drastically over time.

• image_size = 1920
  → Adjusted to match the input video resolution (1080×1920); the default was 1024.

🔹 CoTracker

• Checkpoint: cotracker2.pth
• Destination: co-tracker/cotracker/checkpoints/
• Download: https://github.com/facebookresearch/co-tracker/tree/8d364031971f6b3efec945dd15c468a183e58212

• grid_size = 200
• pad_value = 5

🔹 Edge Detection

• LDC Threshold = 0.7
  → All mask values below 0.7 are considered non-edge (i.e., discarded during boundary detection).
• HoughLineP Threshold = 100
  → Threshold used in Probabilistic Hough Transform to detect lines from the extracted edges.

Results

Citation

• (https://doi.org/10.1016/j.engappai.2025.113718)

@article{ko2026computer,
  title={Computer vision-based steel strand tension evaluation},
  author={Ko, Dongyoung and Park, Minsoo and Na, Sangil and Jang, Juyoung and Cho, Yong Kwon and Park, Seunghee},
  journal={Engineering Applications of Artificial Intelligence},
  volume={166},
  pages={113718},
  year={2026},
  publisher={Elsevier}
}

Contributors

Ph.D. Candidate Dongyoung Ko

Professor Minsoo Park

Professor Seunghee Park