Image and Feature Space Based Domain Adaptation for Vehicle Detection

The application of deep learning in the field of object detection has experienced much progress. However, due to the domain shift problem, applying an off-the-shelf detector to another domain leads to a significant performance drop. A large number of ground truth labels are required when using another domain to train models, demanding a large amount of human and financial resources. In order to avoid excessive resource requirements and performance drop caused by domain shift, this paper proposes a new domain adaptive approach to cross-domain vehicle detection. Our approach improves the cross-domain vehicle detection model from image space and feature space. We employ objectives of the generative adversarial network and cycle consistency loss for image style transfer in image space. For feature space, we align feature distributions between the source domain and the target domain to improve the detection accuracy. Experiments are carried out using the method with two different datasets, proving that this technique effectively improves the accuracy of vehicle detection in the target domain.     

Investigation on milling force of thin-walled workpiece considering dynamic characteristics of workpiece

Journal of Mechanical Science and Technology - The idea of integral calculus is used in traditional milling force prediction methods. In the traditional method, the milling cutter is divided into...     

A quarter century of wine pigment discovery

The red grape berry pigments, anthocyanins, were characterized in the early 20th century, but investigations of wine pigments were stymied during that era. The question of their identity was a major challenge for wine chemists. A number of techniques showed that the pigments were polymeric in nature. Some structures were postulated by mid-century based on reactions between anthocyanins and condensed tannin, and later postulated wine pigments were observed in model reactions. Some related compounds were then observed in wine. By the end of the 20th century, the ionization of non-volatiles for mass spectrometry opened the door to these compounds. In addition, a new class of compounds was observed, the pyranoanthocyanins, a product of fermentation and aging metabolites with anthocyanins. These compounds possess the pigment stability to SO₂ and pH change that is characteristic of aged red wine. Aging experiments show a dynamic situation with shifts in the population of pigment classes over time. The very large number of diverse pigments explains why it has been so difficult to answer the century-old question of the structure of wine pigments. Our current understanding is founded on the use of mass spectral analysis using electrospray and related ionization techniques over the last 25 years. Future progress will rely on more sophisticated analysis of this very complex mixture of substances. © 2019 Society of Chemical Industry     

丙烷脱沥青装置调合前沥青的综合利用

在丙烷脱沥青装置加热炉试烧装置自产调合前沥青,通过观察油枪雾化沥青的能力、火嘴的燃烧情况以及燃烧后的烟尘状况,筛选出适合沥青燃烧的油枪,最终实现调合前沥青直接做加热炉燃料、减少重油出厂量的目的。     

A novel method based on convolutional neural networks for deriving standard 12-lead ECG from serial 3-lead ECG

Frontiers of Information Technology & Electronic Engineering - Reconstruction of a 12-lead electrocardiogram (ECG) from a serial 3-lead ECG has been researched in the past to satisfy the need...     

PET的紫外老化性能及机理研究

采用力学、紫外-可见光分析和热失质等测试手段研究了PET改性料的紫外老化性能及老化机理。结果表明:复合助剂的加入能显著提升体系的力学性能保持率,进而提升其紫外老化寿命,提升的幅度随复合助剂中耐紫外剂含量的增加而显著增加。DSC分析表明,随着复合助剂的引入和老化的进行,改性料结晶度增加。文中进一步讨论了PET改性料的紫外老化机理。紫外-可见光分析表明,紫外老化能产生醌类和双醌类物质,进而提升PET改性料的紫外吸收度。     

Evaluation model of green supply chain cooperation credit based on BP neural network

Neural Computing and Applications - More and more enterprises hope to achieve cooperation and win–win. However, many companies often have problems such as insufficient partner credit, which...     

Predicting visual similarity between colour palettes

This work is concerned with the prediction of visual colour difference between pairs of palettes. In this study, the palettes contained five colours arranged in a horizontal row. A total of 95 pairs of palettes were rated for visual difference by 20 participants. The colour difference between the palettes was predicted using two algorithms, each based on one of six colour-difference formulae. The best performance (r² = 0.86 and STRESS = 16.9) was obtained using the minimum colour-difference algorithm (MICDM) using the CIEDE2000 equation with a lightness weighing of 2. There was some evidence that the order (or arrangement) of the colours in the palettes was a factor affecting the visual colour differences although the MICDM algorithm does not take order into account. Application of this algorithm is intended for digital design workflows where colour palettes are generated automatically using machine learning and for comparing palettes obtained from psychophysical studies to explore, for example, the effect of culture, age, or gender on colour associations.