Browsing by Author "Chang, Wei-Lun"

Now showing 1 - 4 of 4

All about Structure: Adapting Structural Information across Domains for Boosting Semantic Segmentation
(2019-01-01)
Chang, Wei-Lun
;
Wang, Hui-Po
;
Peng, Wen-Hsiao
;
Chiu, Wei-Chen
;
交大名義發表
In this paper we tackle the problem of unsupervised domain adaptation for the task of semantic segmentation, where we attempt to transfer the knowledge learned upon synthetic datasets with ground-truth labels to real-world images without any annotation. With the hypothesis that the structural content of images is the most informative and decisive factor to semantic segmentation and can be readily shared across domains, we propose a Domain Invariant Structure Extraction (DISE) framework to disentangle images into domain-invariant structure and domain-specific texture representations, which can further realize image-translation across domains and enable label transfer to improve segmentation performance. Extensive experiments verify the effectiveness of our proposed DISE model and demonstrate its superiority over several state-of-the-art approaches.
Combined micro- and nano-scale surface textures for enhanced near-infrared light harvesting in silicon photovoltaics
(2011-03-04)
Chang, Chia-Hua
;
Yu, Peichen
;
Hsu, Min-Hsiang
;
Tseng, Ping-Cheng
;
Chang, Wei-Lun
As silicon photovoltaics evolve towards thin-wafer technologies, efficient optical absorption for the near-infrared wavelengths has become particularly challenging. In this work, we present a solution that employs combined micro- and nano-scale surface textures to increase light harvesting in the near-infrared for crystalline silicon photovoltaics, and discuss the associated antireflection and scattering mechanisms. The surface textures are achieved by uniformly depositing a layer of indium-tin-oxide nanowhiskers on micro-grooved silicon substrates using electron-beam evaporation. The nanowhiskers facilitate optical transmission in the near-infrared by functioning as impedance matching layers with effective refractive indices gradually varying from 1 to 1.3. Materials with such unique refractive index characteristics are not readily available in nature. As a result, the solar cell with combined textures achieves over 90% external quantum efficiencies for a broad wavelength range of 460-980 nm, which is crucial to the development of advanced thin-substrate silicon solar cells.
Enhanced angular characteristics of indium tin oxide nanowhisker-coated silicon solar cells
(2011-05-09)
Chang, Chia-Hua
;
Hsu, Min-Hsiang
;
Tseng, Ping-Chen
;
Yu, Peichen
;
Chang, Wei-Lun
Omnidirectional and broadband light harvesting is critical to photovoltaics due to the sun's movement and its wide spectral range of radiation. In this work, we demonstrate distinctive indium-tin-oxide nanowhiskers that achieve superior angular and spectral characteristics for crystalline silicon solar cells using angle-resolved reflectance spectroscopy. The solar-spectrum weighted reflectance is well below 6% for incident angles of up to 70 and for the wavelength range between 400nm and 1000nm. As a result, the nanowhisker coated solar cell exhibits broadband quantum efficiency characteristics and enhanced short-circuit currents for large angles of incidence. (C) 2011 Optical Society of America
Similarity Based ART 1 Model for Automatic Die Defect Detection and Classification
(2014)
Chang, Jyh-Yeong
;
Chang, Wei-Lun
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
In general, a reliable and automatic semiconductor fabrication processes is of great importance to product yield and cost reduction. In the past, we made use of human vision to do wafer die defect detection and classification, which is hindered by the easy fatigue and fuzziness of human eyes and the decision difference between inspectors. In this paper, we develop a vision-based automatic defect classification system. In our defect detection component, we apply the MAD method to align the test image with the reference image. To acquire the binary defect images, we subtract the test image from the reference image, then we convert the difference image into the binary image by setting a threshold. Moreover, we removed the scattering noises by setting a minimum number of connected noisy pixels required. Finally, we extract all defects in the test image in order to perform the defect classification. For defect classification, we revise the ART 1 model, which still can retain its stability and the plasticity dilemma. We have found that ART 1 shows an intolerable shortcoming: whose output is dependent on the ordering of input sequence applied to the ART 1 model. To remedy this disadvantage, we derive the similarity based ART 1 model which can obtain high classification accuracy and independent on the ordering of input patterns applied.