Container transport is widely used in logistics, and understanding the movement patterns of container trailers can improve logistics efficiency. Despite the availability of vehicle GPS data, studies focusing on container trailers in port areas are still limited. This study uses ETC2.0 data to extract trailer trajectories that visited Sendai Port in northern Japan. This port is of moderate size, allowing for the collection of a substantial amount of data, yet not so extensive that it precludes detailed analysis of individual cases. Analysis of the stop points of the trailers entering the container terminal shows similar characteristics to those observed in bigger Japanese ports. Since detecting stops alone may not capture visits to locations relevant to understanding trailer mobility, this study proposes a method to detect even short visits to such places. This method enabled the characterization between trailers transporting empty containers and those carrying loaded containers to the container terminal by the locations visited before and after accessing the container terminal.

In this paper, we propose a method for predicting the impression evaluation of buildings’ exterior appearance using street image big data, and we demonstrate its applicability to architectural design. First, we conduct a large-scale impression evaluation web questionnaire using building exterior images extracted mechanically from street image big data in Ota Ward, Tokyo. Next, by training a deep learning model using the results, we can quantify the impression evaluation of building exterior images taken from different building uses and angles. Furthermore, we analyze the impression evaluation scores predicted by the model and demonstrate this method in architectural design through several case studies.

Remote sensing image semantic segmentation of land use, benefitted from the development of deep learning and consequently made considerable progress in terms of inferencing accuracy and speed. However, the effective training of semantic segmentation models for remote sensing imagery necessitates extensively detailed pixel-level annotations, and gathering such data is both time-intensive and laborious. Thus, this study implemented low-rank adaptation on a stable diffusion algorithm to learn the distribution of the pixel-level annotations in case of the LoveDA dataset. Consequently, the annotation-image pairs were used to train the remote sensing image generator based on stable diffusion guided by ControlNet. We proposed a stable diffusion based approach, which can generate image-annotation pairs from scratch. The generated annotation and image pairs achieved a high accuracy of 0.520 mean intersection-over-union on LoveDA dataset, which is close to the original data training result of 0.539 mIoU. Furthermore, the mixed training using generated and original data achieved 0.542 mIoU, thereby demonstrating the data augmentation function of our approach. This study provided a solution for the high-cost pixel-level annotation issue, and thus, exhibited the potential of artificial intelligence generated content.