Author Affiliations
Abstract
Institute of Photoelectronic Thin Film Devices and Technology of Nankai University, Key Laboratory of Photoelectronic Thin Film Devices and Technology of Tianjin, Engineering Research Center of Thin Film Photoelectronic Technology, Ministry of Education, Tianjin 300350, China
Chalcopyrite (CIGS) thin films deposited in a low-temperature process (450°C) usually produce fine grains and poor crystallinity. Herein, different Ag treatment processes, which can decrease the melting temperature and enlarge band gap of the CIGS films, were employed to enhance the quality of thin films in a low-temperature deposition process. It is demonstrated that both the Ag precursor and Ag surface treatment process can heighten the crystallinity of CIGS films and the device efficiency. The former is more obvious than the latter. Furthermore, the Urbach energy is also reduced with Ag doping. This work aims to provide a feasible Ag-doping process for the high-quality CIGS films in a low-temperature process.
CuSe2 thin film low-temperature deposition process Ag doping crystallinity Urbach energy Chinese Optics Letters
2021, 19(11): 114001
南京工业大学电气工程与控制科学学院， 江苏 南京 211816
在人脸表情识别任务中，传统的机器学习方法是基于人工来提取特征，其特征提取过程时间复杂度高且稳健性差，而现有依赖单通道卷积核的卷积神经网络提取特征不够充分，进而导致识别率不高。针对这些问题，提出一种多分辨率特征融合的卷积神经网络。利用两个相互独立且深度不同的通道对图片进行特征提取，使卷积神经网络自主学习同一图像下不同分辨率的特征，然后将不同分辨率的特征送入全连接层并进行特征融合，最后经过softmax分类器进行表情分类。在JAFFE和CK+表情数据库上进行了多次实验，结果表明，与传统的机器学习方法和现有的卷积神经网络结构相比，所提卷积神经网络结构模型具有稳健性好、泛化能力强、收敛速度快的优点。
机器视觉 人脸表情识别 特征提取 卷积神经网络 多分辨率特征融合 激光与光电子学进展
2018, 55(7): 071503
