Search

Fine-grained recognition task deals with sub-category classification problem, which is important for real-world applications. In this work, we are particularly interested in the segmentation task on the \emph{finest-grained} level, which is specifically named "individual segmentation". In other words, the individual-level category has no sub-category under it. Segmentation problem in the individual level reveals some new properties, limited training data for single individual object, unknown background, and difficulty for the use of depth. To address these new problems, we propose a "Context Less-Aware" (CoLA) pipeline, which produces RGB-D object-predominated images that have less background context, and enables a scale-aware training and testing with 3D information. Extensive experiments show that the proposed CoLA strategy largely outperforms baseline methods on YCB-Video dataset and our proposed Supermarket-10K dataset. Code, trained model and new dataset will be published with this paper., We found some significant errors which could influence the correctness of this paper, and we still need more time to find our solutions

Ammonia (NH3) is an excellent hydrogen (H2) carrier that is easy to bulk manufacture, handle, transport, and use. NH3 is itself combustible and could potentially become a clean transport fuel for direct use in internal combustion engines (ICEs). This technical review examines the current state of knowledge of NH3 as a fuel in ICEs on its own or in mixtures with other fuels. A particular case of interest is to partially dissociate NH3 in situ to produce an NH3/H2 mixture before injection into the engine cylinders. A key element of the present innovation, the presence of H2 is expected to allow easy control and enhanced performance of NH3 combustion. The key thermochemical properties of NH3 are collected and compared to those of conventional and alternative fuels. The basic combustion characteristics and properties of NH3 and its mixtures with H2 are summarized, providing a theoretical basis for evaluating NH3 combustion in ICEs. The combustion chemistry and kinetics of NH3 combustion and mechanisms of NOx formation and destruction are also discussed. The potential applications of NH3 in conventional ICEs and advanced homogenous charge compression ignition (HCCI) engines are analyzed.

Taking the axial exhaust condenser of one gas-steam combined cycle unit as the research object, the steam flow in the shell side of the condenser was simulated under summer operating conditions. The flow distribution of the condenser in the shell side was obtained. Through the extraction of simulation data, the pressure of the condenser was obtained. The simulation results of single factor influence showed that the main factors leading to high operating pressure of the condenser under summer working condition were steam load, circulating water inlet temperature, circulating water flow and vacuum tightness.

The uniform and compact n-type TiO 2 window layers were prepared by RF sputtering using a TiO 2 ceramic target. A high open circuit voltage of about 1.03 V with the efficiency of 13.23% can be obtained as compared to 0.92 V of solution based TiO 2 . Nearly 100% coverage of perovskite film on sputtered TiO 2 can be observed in the scanning electron microscopy images, obviously higher than 92% coverage on solution-based TiO 2 . The full coverage of perovskite and the proper band alignment between sputtered TiO 2 and spiro-OMeTAD are two reasons responsible for the high open circuit voltage. The effects of annealing, growth temperature, sputtering target and thickness on performance of solar cell are also discussed.

The surface texture with a long alignment was observed on the surface of CH3NH3PbI3−xClx films prepared using one-step solution deposition. We found that it appears only when hydroiodic acid with stabilizer H3PO2 is adopted to prepare CH3NH3I powder. However, even with this stabilizer, the surface texture of CH3NH3PbI3−xClx films can also be eliminated completely by sufficiently lateral diffusion before annealing, which can be achieved by high speed coating above 3000 rpm or drying for 30 min at room temperature. The later process results in a smooth surface with homogeneous grains and hence the improvement of photovoltaic performance.

Instance segmentation is a problem of significance in computer vision. However, preparing annotated data for this task is extremely time-consuming and costly. By combining the advantages of 3D scanning, reasoning, and GAN-based domain adaptation techniques, we introduce a novel pipeline named SRDA to obtain large quantities of training samples with very minor effort. Our pipeline is well-suited to scenes that can be scanned, i.e. most indoor and some outdoor scenarios. To evaluate our performance, we build three representative scenes and a new dataset, with 3D models of various common objects categories and annotated real-world scene images. Extensive experiments show that our pipeline can achieve decent instance segmentation performance given very low human labor cost., ECCV 2018