Related Experiment Video
Updated: Aug 5, 2026
08:25
Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
Published on: May 7, 2019
HODet:Horizontal-to-Oriented Object Detection with simplified annotation
Jinlin Chen1,2, Yiquan Wu3, Yubin Yuan1
1College of Electronic and Information Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing, China.
Scientific Reports
|August 4, 2026
Summary
This study introduces a new AI framework for reading ship draft weighing by improving oriented object detection. The method enhances accuracy by refining bounding boxes and reducing manual annotation needs, boosting water gauge reading precision by 35.3%.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Machine Learning
Background:
- Accurate ship draft weighing relies on AI systems for reading water gauges.
- Manual annotation of oriented objects for AI training is labor-intensive and error-prone.
- Existing methods struggle with precise recognition and positioning of characters on water gauges.
Purpose of the Study:
- To develop a novel oriented detection framework for improved AI-based ship draft weighing.
- To reduce the workload and enhance the accuracy of object annotation in datasets.
- To improve the performance of AI systems in recognizing and positioning characters on water gauges.
Main Methods:
- A novel oriented detection framework is proposed, refining horizontal bounding boxes with three additional parameters.
- Two symmetric functions are used to constrain angles and scales within defined ranges.
- A residual network is integrated into the YOLO backbone to enhance feature extraction and small object sensitivity.
- Improved Intersection over Union (IoU) is applied to bounding box regression loss for accurate oriented estimation.
Main Results:
- The framework requires only horizontal annotation information, significantly reducing annotation workload.
- State-of-the-art performance was achieved on aerial object datasets with minimal impact on detection speed.
- Real-time testing on 120 drone videos showed a 35.3% improvement in water gauge reading accuracy.
Conclusions:
- The proposed oriented detection framework effectively addresses challenges in AI-based ship draft weighing.
- The method offers a practical solution for accurate and efficient recognition of oriented objects, particularly in aerial imagery.
- This advancement has significant implications for maritime logistics and safety through improved automated readings.