Search

Bread is one of the most consumed foods in the world, and modern food processing technologies using artificial intelligence are crucial in providing quality control and optimization of food products. An integrated solution of sensor data and machine learning technology was determined to be appropriate for identifying real-time changing environmental variables and various influences in the baking process. In this study, the Baking Process Prediction Model (BPPM) created by data-based machine learning showed excellent performance in monitoring and analyzing real-time sensor and vision data in the baking process to predict the baking stages by itself. It also has the advantage of improving the quality of bread. The volumes of bread made using BPPM were 127.54 ± 2.54, 413.49 ± 2.59, 679.96 ± 1.90, 875.79 ± 2.46, and 1260.70 ± 3.13, respectively, which were relatively larger than those made with fixed baking time (p < 0.05). The developed system is evaluated to have great potential to improve precision and efficiency in the food production and processing industry. This study is expected to lay the foundation for the future development of artificial intelligence and the food industry. [ABSTRACT FROM AUTHOR]

The powder packing behaviour of Fe-based amorphous powders was experimentally monitored and compared with values predicted using the Desmond model and a simulation based on the discrete element method. Powders with different sizes were mixed in various ratios to produce a tri-modal distribution. The simulation results revealed the cohesive force of the powder in terms of the angle of repose and were in agreement with the experimental results. However, the packing behaviour derived from the theoretical model deviated from the experimental results because the interaction among the powder was not considered in the former. Furthermore, the packing fractions for different mixing ratios were investigated through an artificial intelligence framework to determine the optimal mixing ratio. This ratio was validated experimentally and determined to be approximately 8.97% higher than that for the monodisperse large powder case. [ABSTRACT FROM AUTHOR]

Vision and language understanding has emerged as a subject undergoing intense study in Artificial Intelligence. Among many tasks in this line of research, visual question answering (VQA) has been one of the most successful ones, where the goal is to learn a model that understands visual content at region-level details and finds their associations with pairs of questions and answers in the natural language form. Despite the rapid progress in the past few years, most existing work in VQA have focused primarily on images. In this paper, we focus on extending VQA to the video domain and contribute to the literature in three important ways. First, we propose three new tasks designed specifically for video VQA, which require spatio-temporal reasoning from videos to answer questions correctly. Next, we introduce a new large-scale dataset for video VQA named TGIF-QA that extends existing VQA work with our new tasks. Finally, we propose a dual-LSTM based approach with both spatial and temporal attention and show its effectiveness over conventional VQA techniques through empirical evaluations. [ABSTRACT FROM AUTHOR]

We propose an approach for retrieving a sequence of natural sentences for an image stream. Since general users often take a series of pictures on their experiences, much online visual information exists in the form of image streams, for which it would better take into consideration of the whole image stream to produce natural language descriptions. While almost all previous studies have dealt with the relation between a single image and a single natural sentence, our work extends both input and output dimension to a sequence of images and a sequence of sentences. For retrieving a coherent flow of multiple sentences for a photo stream, we propose a multimodal neural architecture called coherence recurrent convolutional network (CRCN), which consists of convolutional neural networks, bidirectional long short-term memory (LSTM) networks, and an entity-based local coherence model. Our approach directly learns from vast user-generated resource of blog posts as text-image parallel training data. We collect more than 22 K unique blog posts with 170 K associated images for the travel topics of NYC, Disneyland , Australia, and Hawaii. We demonstrate that our approach outperforms other state-of-the-art image captioning methods for text sequence generation, using both quantitative measures and user studies via Amazon Mechanical Turk. [ABSTRACT FROM PUBLISHER]

Artificial synapse devices are currently the subjects of great attention as next-generation hardware for data processing to overcome the problem of data explosion due to the rapid advances in artificial intelligence and cloud computing technology. Nanocomposite-based devices enable unique applications and have several advantages that cannot be achieved in single material-based devices. This study presents binary electrical synapses with digital data storing and analog data processing through a nanocomposite-based active layer composed of poly(methyl methacrylate) (PMMA) with embedded chlorine-functionalized graphene quantum dots (fGQDs) on an indium tin oxide (ITO) substrate. The Al/PMMA-fGQD/ITO devices with an fGQD concentrations of 5 wt% exhibited excellent memory performance with R ON /R OFF ratio of 103. Moreover, we demonstrated that our device can successfully emulate biological synaptic functions such as potentiation/depression, short-term/long-term memory, paired-pulse facilitation, learning experience, and spike-timing-dependent plasticity. Furthermore, on the basis of the synaptic behaviors of the devices, they achieved about a 90 % recognition capability when a learning algorithm was used in a single-layer neural network. [Display omitted] • A chlorine-functionalized graphene quantum dots (fGQDs) was introduced as tapping centers of an active layer in an artificial synapse. • The Al/PMMA-fGQDs/ITO device showed excellent electrical characteristics and successfully mimicked various synaptic functions. • Based on the conductance of the Al/PMMA-fGQDs/ITO device, the MNIST pattern recognition algorithm achieved about a 90 % recognition rate. [ABSTRACT FROM AUTHOR]