Abstract:Deep learning-based object detection methods are utilized for safety management at construction sites, which require large-scale, high-quality, and well-labeled datasets for training. The existing construction datasets are relatively small due to the high expense of labor-intensive annotation, and the varying quality of the construction images also affects the detection performance of the model. To address the limitations of datasets, this study proposes a new method for construction object detection by integrating super-resolution and semi-supervised learning. The proposed method improves the quality of construction images and achieves excellent detection performance with limited labeled data. First, the Real-ESRGAN model is introduced to improve the quality of construction images and make the construction objects visible. The proposed super-resolution method can enhance the texture details of low-resolution images, hence improving the performance of object detection models. Second, the mean-teacher network is adopted to expand the training set, thus avoiding the labor-intensive annotation work. To verify the effectiveness of the proposed method, the method is applied to the state-of-the-art Yolov5 object detection model, and construction images from the Site Object Detection Dataset (SODA) with different labeled data proportions (from 10% to 50% in 10% intervals with an extreme case of 5%) are used as the training set. By comparing with the existing supervised learning method, it is shown that the proposed method can achieve better detection performance. In particular, the method is more effective in enhancing detection performance when the proportion of the labeled data is smaller, which is of great practical value in real-world engineering. The experimental results show the potential of the proposed method in improving image quality and reducing the expense of developing construction datasets.

Semi-supervised symbol detection for piping and instrumentation drawings

OSSR-PID: One-Shot Symbol Recognition in P&ID Sheets using Path Sampling and GCN

Features Recognition from Piping and Instrumentation Diagrams in Image Format Using a Deep Learning Network

Semi-supervised learning approach for construction object detection by integrating super-resolution and mean teacher network

Digitize-PID: Automatic Digitization of Piping and Instrumentation Diagrams

Automatic Information Extraction from Piping and Instrumentation Diagrams

Automatic Detection and Classification of Symbols in Engineering Drawings

Automatic anomaly detection in engineering diagrams using machine learning

Deep learning for symbols detection and classification in engineering drawings

Few-Shot Symbol Detection in Engineering Drawings

Symbols in Engineering Drawings (SiED): An Imbalanced Dataset Benchmarked by Convolutional Neural Networks

Semi-supervised pipeline anomaly detection algorithm based on memory items and metric learning

End-to-end digitization of image format piping and instrumentation diagrams at an industrially applicable level

The Computer Vision-based Tolerancing Callout Detection Model

CISO: Co-iteration Semi-Supervised Learning for Visual Object Detection

Pixel-Wise Symbol Spotting via Progressive Points Location for Parsing CAD Images

An Efficient Primitive-Based Method to Recognize Online Sketched Symbols with Autocompletion

Deep Learning-Based Method to Recognize Line Objects and Flow Arrows from Image-Format Piping and Instrumentation Diagrams for Digitization

Drainage Pipeline Defect Detection System Based on Semantic Segmentation

Towards automatic generation of Piping and Instrumentation Diagrams (P&IDs) with Artificial Intelligence

Symbol as Points: Panoptic Symbol Spotting via Point-based Representation