When detecting aircraft flared tube defect with the YOLOv4 network structure, tiny object defects will be missed, and resulting in a high missed detection rate and low mAP (mean Average Precision) value. This paper proposes an improved aircraft flared tube defect detection algorithm of YOLOv4 network structure. Firstly, in order to improve the feature extraction capability of the YOLOv4 network for tiny defects, a convolution operation is added to the SPP (Spatial Pyramid Pooling) and PANet (Path Aggregation Network) structure. Secondly, the representation of the feature pyramid is enhanced utilizing the improved PANet. Thirdly, the decoupled head is utilized to improve the model performance. Finally, we construct the aircraft flared tube dataset by labeling the defect samples, and experiment with the improved YOLOv4 network. Experimental results show that the mAP value of defect detection task is increased from 91.26% to 95.31%, average detection time increased from 346.17ms to 278.61ms.
Nowadays, with the rapid expansion of social media as a means of quick communication, real-time disaster information is widely disseminated through these platforms. Determining which real-time and multi-modal disaster information can effectively support humanitarian aid has become a major challenge. In this paper, we propose novel end-to-end model, named GCN-based Semi-supervised Multi-modal Domain Adaptation (GSMDA), which consists of three essential modules: the GCN-based feature extraction module, the attention-based fusion module, and the MMD domain adaptation module. The GCN-based feature extraction module integrates text and image representations through GCNs, while the attention-based fusion module then merges these multi-modal representations using an attention mechanism. Finally, the MMD domain adaptation module is utilized to alleviate the dependence of GSMDA on source domain events by computing the maximum mean discrepancy across domains. our experiments results demonstrate that GSMDA outperforms the current state-of-the-art models in terms of performance and stability.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.