We present a Multi-Instance Generation (MIG) task, si-multaneously generating multiple instances with diverse controls in one image. Given a set of predefined coordinates and their corresponding descriptions, the task is to ensure that generated instances are accurately at the designated locations and that all instances' attributes adhere to their corresponding description. This broadens the scope of current research on Single-instance generation, elevating it to a more versatile and practical dimension. Inspired by the idea of divide and conquer, we introduce an innovative approach named Multi-Instance Generation Controller (MIGC) to address the challenges of the MIG task. Ini-tially, we break down the MIG task into several subtasks, each involving the shading of a single instance. To ensure precise shading for each instance, we introduce an instance enhancement attention mechanism. Lastly, we aggregate all the shaded instances to provide the necessary information for accurately generating multiple instances in stable diffusion (SD). To evaluate how well generation models per-form on the MIG task, we provide a COCO-MIG bench-mark along with an evaluation pipeline. Extensive experiments were conducted on the proposed COCO-MIG bench-mark, as well as on various commonly used benchmarks. The evaluation results illustrate the exceptional control ca-pabilities of our model in terms of quantity, position, at-tribute, and interaction. Code and demos will be released at https://migcproject.github.io/.

Online monitoring of laser welding depth is increasingly important, with the growing demand for the precise welding depth in the field of power battery manufacturing for new energy vehicles. The indirect methods of welding depth measurement based on optical radiation, visual image and acoustic signals in the process zone have low accuracy in the continuous monitoring. Optical coherence tomography (OCT) provides a direct welding depth measurement during laser welding and shows high achievable accuracy in continuous monitoring. Statistical evaluation approach accurately extracts the welding depth from OCT data but suffers from complexity in noise removal. In this paper, an efficient method coupled DBSCAN (Density-Based Spatial Clustering of Application with Noise) and percentile filter for laser welding depth determination was proposed. The noise of the OCT data were viewed as outliers and detected by DBSCAN. After eliminating the noise, the percentile filter was used to extract the welding depth. By comparing the welding depth determined by this approach and the actual weld depth of longitudinal cross section, an average error of less than 5% was obtained. The precise laser welding depth can be efficiently achieved by the method.

The reported incidence of scoliosis among adolescents in China differs according to screening method owing to the lack of uniformity and limitations of certain techniques. We aimed to design, develop, and validate a non-invasive, accurate, portable, fast, and automated tool that would enable the measurement and storage of data during scoliosis screening. We designed a new portable electronic scoliosis screening device (PESSD)—for the identification of adolescent scoliosis based on ergonomics theory. The device measured the axial deflection angle of the trunk of the human body using a built-in angle sensor. Data obtained using the PESSD, a traditional scoliometer manual ruler, and X-ray measurement of the Cobb angle were compared. The PESSD exhibited more sensitive detection of small-angle scoliosis and improved repeatability compared with the scoliometer. The data obtained using the PESSD showed good correlation with Cobb angle data measured from X-ray images. All patients who were indicated to be positive for scoliosis using the PESSD were found to have clinically identifiable scoliosis from X-ray examination. The PESSD may be able to achieve early detection of scoliosis in adolescents. It is non-invasive, highly precise, portable, easy to use, and offers automated data storage and traceability. This study is a pilot or preliminary validation study. With further, more in depth studies, the PESSD has excellent potential for transformation into an effective tool for use in large-scale screening programs for adolescent scoliosis in schools and communities. This article is about designing a new portable electronic scoliosis screening device based on ergonomics theory. Because there are currently no uniform screening methods and standards, the results in this article could facilitate the adoption of a uniform screening tool into large-scale screening programs for adolescent scoliosis in schools and communities, preliminary examination in hospitals, and self-testing at home after parent training.

Vision–language navigation (VLN) is a challenging task, which guides an agent to navigate in a realistic environment by natural language instructions. Sequence-to-sequence modeling is one of the most prospective architectures for the task, which achieves the agent navigation goal by a sequence of moving actions. The line of work has led to the state-of-the-art performance. Recently, several studies showed that the beam-search decoding during the inference can result in promising performance, as it ranks multiple candidate trajectories by scoring each trajectory as a whole. However, the trajectory-level score might be seriously biased during ranking. The score is a simple averaging of individual unit scores of the target-sequence actions, and these unit scores could be incomparable among different trajectories since they are calculated by a local discriminant classifier. To address this problem, we propose a global normalization strategy to rescale the scores at the trajectory level. Concretely, we present two global score functions to rerank all candidates in the output beam, resulting in more comparable trajectory scores. In this way, the bias problem can be greatly alleviated. We conduct experiments on the benchmark room-to-room (R2R) dataset of VLN to verify our method, and the results show that the proposed global method is effective, providing significant performance than the corresponding baselines. Our final model can achieve competitive performance on the VLN leaderboard.

Space exploration missions involve significant participation from astronauts. Therefore, it is of great practical importance to assess the astronauts’ performance via various parameters in the cramped and weightless space station. In this paper, we proposed a calibration-free multi-view vision system for astronaut performance capture, including two modules: (1) an alternating iterative optimization of the camera pose and human pose is implemented to calibrate the extrinsic camera parameters with detected 2D keypoints. (2) Scale factors are restricted by the limb length to recover the real-world scale and the shape parameters are refined for subsequent postural reconstruction. These two modules can provide effective and efficient motion capture in a weightless space station. Extensive experiments using public datasets and the ground verification test data demonstrated the accuracy of the estimated camera pose and the effectiveness of the reconstructed human pose.