Real-Time Ground-Level Building Damage Detection Based on Lightweight and Accurate YOLOv5 Using Terrestrial Images

Abstract

Real-time building damage detection effectively improves the timeliness of post-earthquake assessments. In recent years, terrestrial images from smartphones or cameras have become a rich source of disaster information that may be useful in assessing building damage at a lower cost. In this study, we present an efficient method of building damage detection based on terrestrial images in combination with an improved YOLOv5. We compiled a Ground-level Detection in Building Damage Assessment (GDBDA) dataset consisting of terrestrial images with annotations of damage types, including debris, collapse, spalling, and cracks. A lightweight and accurate YOLOv5 (LA-YOLOv5) model was used to optimize the detection efficiency and accuracy. In particular, a lightweight Ghost bottleneck was added to the backbone and neck modules of the YOLOv5 model, with the aim to reduce the model size. A Convolutional Block Attention Module (CBAM) was added to the backbone module to enhance the damage recognition effect. In addition, regarding the scale difference of building damage, the Bi-Directional Feature Pyramid Network (Bi-FPN) for multi-scale feature fusion was used in the neck module to aggregate features with different damage types. Moreover, depthwise separable convolution (DSCONV) was used in the neck module to further compress the parameters. Based on our GDBDA dataset, the proposed method not only achieved detection accuracy above 90% for different damage targets, but also had the smallest weight size and fastest detection speed, which improved by about 64% and 24%, respectively. The model performed well on datasets from different regions. The overall results indicate that the proposed model realizes rapid and accurate damage detection, and meets the requirement of lightweight embedding in the future.

Disaster impact analysis in Africa: EDDA - UAV building damage dataset of rural and urban Mozambique

“…To test this we use the well known Yolo object detection model which has in recent years been successfully used in many remote sensing detection tasks. 11,14,15,44 For the purpose of this work the Yolo-v7 4 is used.…”

“…Therefore, leveraging the machine vision techniques for this purpose has been largely explored, with the community unanimously transitioning from classical image processing methods 12,13 to learning based approaches. 11,[14][15][16] The literature on the topic has become very rich in the recent years. Therefore, [17][18][19][20] introduced reviews on the topic of structural damage assessment from different perspectives.…”

Deep Learning Based Defect Detection Method for Overhead Transmission Wires

“…In order to verify that the HorCM_PAM_YOLOv5 algorithm proposed in this paper is more suitable to be applied to overhead transmission conductor defect detection scenarios compared to other models, the proposed algorithm is compared with various mainstream algorithms, including the base algorithm YOLOv5s, the base attention mechanism CBAM_YOLOv5s [31], YOLOv6 [32], YOLOv7 [33] and YOLOv8 [34]. The effects are shown in Table 4.…”

Lightweight semantic segmentation of complex structural damage recognition for actual bridges

“…Jiang et al 38 proposed a lightweight object detection network for real-time multi-type structural damage localization, and a YOLOv5-based object detection model was designed for ground-level building seismic damage localization. 39 A lightweight DenseNet with thin autoencoders, 40 depth-separable-convolution enhanced SegNet, 41 and super-resolution-based encoder-decoder with few convolutional parameters 42 have been investigated for bridge crack segmentation. In addition to structural damage recognition, lightweight models are applied to pavement distress classification, 43 tunnel defect segmentation, 44 and road surface crack segmentation.…”