Learning Single-Shot Detector with Mask Prediction and Gate Mechanism

Detection efficiency plays an increasingly important role in object detection tasks.One-stage methods are widely adopted in real life because of their high efficiency especially in some real-time detection tasks such as face recognition and self-driving cars.RetinaMask achieves significant progress in the field of one-stage detectors by adding a semantic segmentation branch,but it has limitation in detecting multi-scale objects.To solve this problem,this paper proposes RetinaMask with Gate(RMG)model,consisting of four main modules.It develops RetinaMask with a gate mechanism,which extracts and combines features at different levels more effectively according to the size of objects.It firstly extracted multi-level features from input image by ResNet.Secondly,it constructed a fused feature pyramid through feature pyramid network,then gate mechanism was employed to adaptively enhance and integrate features at various scales with the respect to the size of object.Finally,three prediction heads were added for classification,localization and mask prediction,driving the model to learn with mask prediction.The predictions of all levels were integrated during the post-processing.The augment network shows better performance in object detection without the increase of computation cost and inference time,especially for small objects.

The Coincidence of Critical Day Length Recognition for Florigen Gene Expression and Floral Transition under Long-Day Conditions in Rice

The photoperiodic control of flowering time is essential for the adaptation of plants to variable environments and for successful reproduction. The identification of genes encoding florigens, which had been elusive but were supposedly synthesized in leaves and then transmitted to shoot apices to induce floral transitions, has greatly advanced our understanding of the photoperiodic regulation of flowering. Studies on the photoperiodism of Arabidopsis, a model long-day plant, revealed the molecular mechanisms regulating the expression of the Arabidopsis florigen gene FT, which is gradually induced in response to increase in day length. By contrast, in rice, a model short-day plant, the expression of the florigen gene Hd3a （an FTortholog in rice） is regulated in an on/off fashion, with strong induction under short-day conditions and repression under long-day conditions. This critical day length dependence of Hd3a expression enables rice to recognize a slight change in the photoperiod as a trigger to initiate floral induction. Rice possesses a second florigen gene, RFT1, which can be expressed to induce floral transition under non-inductive long-day conditions. The complex transcriptional regulation of florigen genes and the resulting precise control over flowering time provides rice with the adaptability required for a crop species of increasing global importance.

Numerical simulation of turbulent flow behind sluice gate under submerged discharge conditions

The turbulent flow discharging under a sluice gate is one kind of typical flows with complicated boundaries in hydraulic and hydroelectric engineering, and the characteristics of its flow field and flow evolution are important in engineering hydraulics. Although there were many studies in this regard, which mainly focus on the discharge capacity, the local scour and the mean flow field, some issues remain to be further investigated, for example, the variation of the mechanical energy during the scouring process, and the adjustment of the unsteady turbulent flow after the local scour with the downstream steady uniform flow. In this paper, the turbulent flow behind a sluice gate is divided into a rapidly varied flow region and a gradually varied flow region, and the above problems are investigated by theoretical analysis and numerical simulation. The main conclusions are as follows：（1） In the discharging process of the turbulent flow under a sluice gate, the river bed would be scoured continuously and the water–air interface is adjusted accordingly, which leads to a decrease of the mechanical energy of the total flow in the rapidly varied flow region, and the mechanical energy loss would also decrease with the increase of the time, yet the variation of the flow discharge under the sluice gate is very small.（2） In the link between the rapidly varied flow and the uniform flow downstream, the mean wall shear stress and the coefficient for the mechanical energy loss would decrease in the longitudinal direction, the decay of the turbulent kinetic energy near the free surface is much more significant than that near the bed in the longitudinal direction, and the mean turbulent kinetic energy in the section would decrease in the longitudinal direction and the decreasing rate is smaller than that of the mean wall shear stress.

Fault Information Recognition for On-board Equipment of High-speed Railway Based on Multi-neural Network Collaboration

It is of great significance to guarantee the efficient statistics of high-speed railway on-board equipment fault information,which also improves the efficiency of fault analysis. Considering this background, this paper presents an empirical exploration of named entity recognition(NER) of on-board equipment fault information. Based on the historical fault records of on-board equipment, a fault information recognition model based on multi-neural network collaboration is proposed. First, considering Chinese recorded data characteristics, a method of constructing semantic features and additional features based on character granularity is proposed. Then, the two feature representations are concatenated and passed into the gated convolutional layer to extract the dependencies from multiple different subspaces and adjacent characters in parallel. Next, the local features are transmitted to the bidirectional long short-term memory(BiLSTM) to learn long-term dependency information. On top of BiLSTM, the sequential conditional random field(CRF) is used to jointly decode the optimized tag sequence of the whole sentence. The model is tested and compared with other representative baseline models. The results show that the proposed model not only considers the language characteristics of on-board fault records, but also has obvious advantages on the performance of fault information recognition.

Defectivity control of aluminum chemical mechanical planarization in replacement metal gate process of MOSFET

The replacement metal gate（RMG） defectivity performance control is very challenging in high-k metal gate（HKMG） chemical mechanical polishing（CMP）. In this study, three major defect types, including fall-on particles, micro-scratch and corrosion have been investigated. The research studied the effects of polishing pad,pressure, rotating speed, flow rate and post-CMP cleaning on the three kinds of defect, which finally eliminated the defects and achieved good surface morphology. This study will provide an important reference value for the future research of aluminum metal gate CMP.