Depth estimation using monocular camera sensors is an important technique in computer vision. Supervised monocular depth estimation requires a lot of data acquired from depth sensors. However, acquiring depth data is an expensive task. We sometimes cannot acquire data due to the limitations of the sensor. View synthesis-based depth estimation research is a self-supervised learning method that does not require depth data supervision. Previous studies mainly use the convolutional neural network (CNN)-based networks in encoders. The CNN is suitable for extracting local features through convolution operation. Recent vision transformers (ViTs) are suitable for global feature extraction based on multiself-attention modules. In this article, we propose a hybrid network combining the CNN and ViT networks in self-supervised learning-based monocular depth estimation. We design an encoder–decoder structure that uses CNNs in the earlier stage of extracting local features and a ViT in the later stages of extracting global features. We evaluate the proposed network through various experiments based on the Karlsruhe Institute of Technology and Toyota Technological Institute (KITTI) and Cityscapes datasets. The results showed higher performance than previous studies and reduced parameters and computations. Codes and trained models are available at https://github.com/fogfog2/manydepthformer .

Originally introduced by H. A. Pohl in 1951, dielectrophoretic (DEP) force has been used as a striking tool for biological particle manipulation (or separation) for the last few decades. In particular, dielectrophoresis activated cell sorters (DACSes) have been developed for applications in various biomedical fields. These applications include cell replacement therapy, drug screening and medical diagnostics. Since a DACS does not require a specific bio-marker, it is able to function as a biological particle sorting tool with numerous configurations for various cells [e.g. red blood cells (RBCs), white blood cells (WBCs), circulating tumor cells, leukemia cells, breast cancer cells, bacterial cells, yeast cells and sperm cells]. This article explores current DACS capabilities worldwide, and it also looks at recent developments intended to overcome particular limitations. First, the basic theories are reviewed. Then, representative DACSes based on DEP trapping, traveling wave DEP systems, DEP field-flow fractionation and DEP barriers are introduced, and the strong and weak points of each DACS are discussed. Finally, for the purposes of commercialization, prerequisites regarding throughput, efficiency and recovery rates are discussed in detail through comparisons with commercial cell sorters (e.g. fluorescent activated and magnetic activated cell sorters).

This article presents a high precise deployment mechanism for a deployable space telescope to facilitate satellite miniaturization. It is designed with a passive deployment mechanism utilizing a spring hinge. In particular, the customized modules and an assembly jig are specifically designed to reduce alignment errors. To confirm the feasibility of the designed mechanism, three alignment errors that influence the optical performance of the structure—tilt, de-center, and de-space—are theoretically analyzed for quarter, half, and full model, respectively. In the case of quarter model, significant results are obtained as a tilt of 21.12 µrad, a de-center of 2.20 µm, and a de-space of 1.71 µm. Based on the theoretical analysis, a deployment mechanism of the quarter model is fabricated, and the alignments of the deployment mechanism are experimentally investigated with a measurement platform consisting of five non-contact-based laser displacement sensors. In addition, the influence of gravity on the alignment error is analyzed and compensated by investigating the tendency of the alignment error according to the rotation degree of the measurement model for the direction of gravity. As a result of the gravity compensation, the proposed mechanism gives acceptable alignment errors as a tilt of 30.04 µrad, de-center of 8.92 µm, and de-space of 4.03 µm, which are controllable by employing the conventional focusing mechanism.