Search

Constructing an efficient photocatalyst for H2 production generally requires the integration of several functional features in and on a semiconductor platform. This is conventionally done by a classic step-by-step synthesis. Here, we report the self-assembly of an amorphous titania nanotube based system when illuminated for extended time in suitable aqueous solutions. Over time, not only photodeposition of a co-catalyst can take place, but also in situ crystallization and morphology changes occur that can provide a higher H2 production than observed for a classically assembled photocatalyst.

This paper presents an improved traffic sign detection method based on Faster R-CNN with dataset augmentation and subcategory detection scheme. Firstly, we extract natural scene frames from given videos and determine 20 categories of traffic signs. Secondly, we extend the image dataset and extract regions of interest, then manually annotate all categories. Thirdly, we train the Faster R-CNN model based on TensorFlow, then test the model and obtain the following evaluation indexes: the mean average precision is 99.07%, the recall rate is 99.66%, and the precision rate is 97.54%. Finally, we add the subcategory detection scheme to determine traffic light states, and we get the following evaluation indexes: the mean average precision is 99.50\%, the recall rate is 100%, and the precision rate is 94.40\%. Our experiments prove the robustness and accuracy for both traffic sign detection and subcategory detection of traffic light.

An updated framework based on LDA is provided in this paper to extract features from online user reviews which are in Chinese. This model is an extension of the LDA by introducing the concepts of multi-gram and part of speech into it and it is named PM-LDA. Through this model, features are generated as topics and topic labels can be generated as the sentence that has the max topic probability. Topics in PM-LDA are divided into two different types. The one is some general features such as product brand, color or producing area, and the other is those latent characteristics which customers may be more interested in. Part of speech is used to get the feature object and feature opinion separately, to make it more accurate. Several experiments are carried out to help to evaluate this model and it is indicated that this method has improved performance both in accuracy and the quality of the topic model itself.