Beach wrack is a common phenomenon along global coastlines. A certain part of macrophytes detached from Blue carbon ecosystems (BCEs) is washed up onshore and accumulated as beach wrack. As one of the largest amounts of production exported from BCEs, beach wrack serves a significant role in the carbon cycling of BCEs. However, compared with the estimation of buried production in BCEs, evaluating the amount of this exported part remains challenging due to the difficulty in tracking them under highly dynamic coastal environments. Thus, we employed Unmanned aerial vehicles (UAVs) and camera traps, which could acquire higher spatial-temporal images than traditional satellites, to investigate the deposition and relocation of beach wrack. Machine learning (ML) was applied to analyze the achieved image datasets. For the aerial images acquired by the UAV, three typical ML methods of K-Nearest Neighbor (KNN), Support Vector Machine (SVM), and Random Forests (RF) performed well in classifying the images (overall classification accuracy > 75 %). Moreover, these ML methods showed geographical transferability to beaches with various characteristics. For the camera images acquired by the camera trap, the deep learning method with a VGG network showed the capability to recognize beach wrack from images recording different beach scenes. Furthermore, compared with manual recognition (20 mins for one image), the deep learning method provided a labor-saving way to analyze large image datasets (5 mins for 187 images). Our study demonstrated that UAVs and camera traps could be applied as labor- and cost-saving tools to conduct customized monitoring tasks to track the dynamic change of beach wrack. The application of ML in classifying image datasets here also revealed the possibility of using large image datasets to explore dynamic ecosystems.