Multitype Damage Detection of Container Using CNN Based on Transfer Learning

Abstract

Due to the repeated bearing of mechanical operations and natural factors, the container will suffer various types of damage during use. Adopting effective container damage detection methods plays a vital role in prolonging the service life and using function. This paper proposes a multitype damage detection model for containers based on transfer learning and MobileNetV2. In addition, a data set containing nine typical types of container damage is established. To ensure the validity and practicability of the model, we conducted tests and verifications in the actual port environment. The results show that the model can identify multiple types of container damage. Compared with the existing models, the damage detection model proposed in this paper can ensure the identification effect of various types of container damage, which is more suitable for the actual container detection situation. This method can provide a new idea of damage detection for container management in ports.

Methodology for generating synthetic labeled datasets for visual container inspection

“…They used different models like Faster R-CNN, SSD-MobileNet and SSD Inception V2 and added an anchor box optimizer in order to detect and localize the corrosion. Classification is also present in damage classification, using a lightweight CNN (MobileNetv2) based on transfer learning, Wang et al (Wang et al, 2021) were able to deploy the model in mobile phone devices. In addition, an end-to-end architecture that faces globally the visual inspection process of containers, grouping and solving multiple visual inspection tasks has been presented (Delgado, 2022).…”

“…YOLOv7 architecture is an extension of YOLOv4 (Bochkovskiy et al, 2020), Scaled-Yolov4 (Wang et al, 2021) and YOLOR by adding several architectural reforms as an Extended Efficient Layer Aggregation Network (E-ELAN) and a trainable bag of freebies. E-ELAN is the computational block in the YOLOv7 backbone, and it is designed to improve speed and accuracy.…”

“…CM R-CNN (Cai and Vasconcelos, 2019) 74.7 13.8 53.9 80.9 Mask R-CNN (He and Gkioxari, 2017) 71.7 13.2 50.9 77.7 MS R-CNN (Huang et al, 2019) 72.9 15.1 51.7 78.9 Faster R-CNN (Ren et al, 2017) 72.7 12.5 52.0 79.1 CRPN Faster R-CNN (Vu et al, 2019) 69.5 15.2 47.7 75.9 Dynamic R-CNN 72.6 19.6 51.7 78.3 Cascade R-CNN (Vu et al, 2019) 74.6 12.8 52.5 82.0 Sparse R-CNN (Sun et al, 2011) 75.8 25.0 55.7 81.2 (He and Gkioxari, 2017) 59.3 18.4 36.5 63.0 MS R-CNN (Huang et al, 2019) 61.1 19.4 39.6 64.4 SOLO 51.5 9.5 27.9 57.0 SOLOv2 56 (Li et al, 2209) 78.0 93.0 84.0 79.0 YOLOv7 79.9 95.3 83.0 79.6 YOLOR-P6 (Wang et al, 2105) 51.2 98.6 77.4 72.2 Scaled-YOLOv4-P7 (Wang et al, 2021) 56.9 96.3 83.7 79.4…”

Detecting Shipping Container Impacts with Vertical Cell Guides inside Container Ships during Handling Operations

“…The inclusion of a broader spectrum of parameters and sensory units could potentially increase the viability of these processes and enable more accurate estimations [ 16 ]. In addition, most of the existing research focuses on damage detection based on visual inspection, using real image datasets produced using visual inspection tools, collected by the repair personnel, related port authorities, and logistics companies [ 17 ]. However, most of the scientific community whose work focuses on damage detection systems for logistics does not have free access to such datasets, as most experiments are forbidden in the port environment for security and legal reasons.…”

An ensemble approach for enhancing generalization and extendibility of deep learning facilitated by transfer learning: principle and application in curing monitoring

“…Lin et al [27] developed a feature extractor that generates damage-sensitive and domain-invariant features, bridging the gap between engineering applications. Wang et al [28] utilized TL to determine damage types caused by the repeated bearing of mechanical operations and natural factors, enhancing the service life of a pressure container. Dunphy et al [29] applied TL and generative adversarial networks for multiclass damage detection within infrastructures.…”