• IEMEK Journal of Embedded Systems and Applications
• /
• v.17 no.4
• /
• pp.217-228
• /
• 2022

Currently, there are increasing demands for applying deep neural networks (DNNs) in the embedded domain such as classification and object detection. The DNN processing in embedded domain often requires custom hardware such as NPU for acceleration due to the constraints in power, performance, and area. Processing DNN models requires a large amount of data, and its seamless transfer to NPU is crucial for performance. In this paper, we developed a cycle-accurate NPU simulator to evaluate diverse NPU microarchitectures. In addition, we propose a novel technique for reducing the number of memory accesses when processing convolutional layers in convolutional neural networks (CNNs) on the NPU. The main idea is to reuse data with memory interleaving, which recycles the overlapping data between previous and current input windows. Data memory interleaving makes it possible to quickly read consecutive data in unaligned locations. We implemented the proposed technique to the cycle-accurate NPU simulator and measured the performance with LeNet-5, VGGNet-16, and ResNet-50. The experiment shows up to 2.08x speedup in processing one convolutional layer, compared to the baseline.

• Smart Media Journal
• /
• v.11 no.11
• /
• pp.25-31
• /
• 2022

Deep learning-based image processing is essential for autonomous vehicles. To process road images in real-time in a System-on-Chip (SoC) environment, we need to execute deep learning models on a NPU (Neural Procesing Units) specialized for deep learning operations. In this study, we imported seven open-source image processing deep learning models, that were developed on GPU servers, to Texas Instrument Deep Learning (TIDL) NPU environment. We confirmed that the models imported in this study operate normally in the SoC virtual environment through performance evaluation and visualization. This paper introduces the problems that occurred during the migration process due to the limitations of NPU environment and how to solve them, and thereby, presents a reference case worth referring to for developers and researchers who want to port deep learning models to SoC environments.

• The Journal of the Acoustical Society of Korea
• /
• v.11 no.6
• /
• pp.41-45
• /
• 1992

음성의 동적인 특성을 이용할 수 있으며, 간단한 알고리즘으로 음성을 인식할 수 있는 NPU 선 형매칭을 이용한 한국어 단어인식에 관하여 연구하였다. 이 인식방법은 NPU 라는 뉴럴 예측기를 적용 한 선형매칭 방법을 이용함으로써, 음성의 동적인 특성을 과거 특징벡터 시계열의 상관관계에 의한 예 측이라는 형태로 인식에 이용하였다. 이 인식방법의 유효성을 확인하기 위해 DDD 지역명을 대상으로 실험한 결과, 96.4%의 인식율을 얻었다.

• Korean Journal of Food Science and Technology
• /
• v.12 no.2
• /
• pp.109-114
• /
• 1980

Protein efficiency ratio (PER), net protein utilization (NPU), nitrogen digestibility (ND), organ weights, and amino acid composition were investigated for rapeseed protein isolates (RPI) prepared by alkaline countercurrent extraction and isoelectric precipitation of defatted rapeseed meal (DRM). The PER values for 3 kinds of RPI obtained at pI of 6.7, 5.6 and 5.0 were 2.8, 3.1 and 2.9, and for DRM 2.5 while the NPU values for these RPI appeared to be 68, 73 and 71 %, and for DRM 56 %, respectively. Mean ND (85%), food intake (87), and weight gain (37.6) for RPI were significantly different from those of DRM (71 %, 77, and 28, respectively). There was no signicant difference in the PER among three RPI and casein (3.0), nor in the NPU among those proteins and casein (74%). These data attributed to the favorable amino acid patterns of proteins isolated which contained balanced essential amino acids in proportions which meet the requirements of human adult(FAO/WHO, 1973).

• ETRI Journal
• /
• v.44 no.5
• /
• pp.849-858
• /
• 2022

Dynamic voltage frequency scaling (DVFS) has been widely adopted for runtime power management of various processing units. In the case of neural processing units (NPUs), power management of neural network applications is required to adjust the frequency and voltage every layer to consider the power behavior and performance of each layer. Unfortunately, DVFS is inappropriate for layer-wise run-time power management of NPUs due to the long latency of voltage scaling compared with each layer execution time. Because the frequency scaling is fast enough to keep up with each layer, we propose a layerwise dynamic frequency scaling (DFS) technique for an NPU. Our proposed DFS exploits the highest frequency under the power limit of an NPU for each layer. To determine the highest allowable frequency, we build a power model to predict the power consumption of an NPU based on a real measurement on the fabricated NPU. Our evaluation results show that our proposed DFS improves frame per second (FPS) by 33% and saves energy by 14% on average, compared with DVFS.

• Nutritional Sciences
• /
• v.3 no.2
• /
• pp.66-70
• /
• 2000

The study was conducted to determine the nutiritional value of Masoor (Lens esculenta) in raw and cooked forms. Supplement value of various types of meat i.e. poultry, mutton and beef at 10, 15, and 20 percent levels for dite containing cooked Masoor was also assessed. Nutritional value of Masoor was determined by chemical analysis as well as through rat assay. Masoor contained an average of 23.18 percent protein and less than two percent fibre. Conventional method of cooking resulted in about 2 per cent increase in Masoor protein. Masoor had 0.83 percent of lysine and cooking destroyed 18 percent of it. Other amino acids in Masoor also showed losses on cooking. Protein efficiency ratio (PER) of diets containing raw Masoor was 1.49 and was reduced to 1.44 by cooking. Cooking of Masoor did not alter true digestibility (TD) percentage. However, net protein utilization (NPU) was improved from 44.60 in raw to 47.77 in cooked. Diets containing cooked Masoor and supplemented with different types of meat significantly improved PER (1.45 to 1.65), TD 76.03 to 87.84 percent and NPU 42.84 to 50.72 percent over non supplemented diets. 20 percent level of supplemented meat showed comparatively better results than other levels in case of improvement in PER, TD and NPU.

• Journal of Nutrition and Health
• /
• v.18 no.3
• /
• pp.234-241
• /
• 1985

The purpose of this study was to investigate changes in nutritive quality of fermented wheat bran prepared by culturing with a microorganism, Aspergillus oryzae, in an attempt to improve the quality of protein in feedstuff. After incubation of wheat bran with Aspergillus oryzae, the contents of chemical composition, including amino acids, riboflavin and amino-nitrogen were increased, but the level of nitrogen free extract was reduced. The effects of supplementation of fermented wheat bran on the rat diets were evaluated by measuring growth rate, feed efficiency and biological values, such as NPU, PER and NPR. Sixty four male Sprague Dawley rats of 5-6 weeks of age were adopted for the feeding trial for 10 days period, and levels of dietary protein were set at 10%. The %contribution of protein from casein, wheat bran and fermented wheat bran for the other dietary treatments were, in the order, 50-50-0%, 50-30-20%, 50-20-30% and 50-0-50%, respectively. In the result of this study, no significant difference were observed in the amount of feed intake body weight gain and feed efficiency. On the whole, the values of NPU, PER and NPR of all the fermented wheat bran groups appeared to be higher than those of the group whose half of the protein was from wheat bran.

• Proceedings of the Korean Institute of Communication Sciences Conference
• /
• 1992.06a
• /
• pp.178-181
• /
• 1992

• Nuclear Engineering and Technology
• /
• v.51 no.8
• /
• pp.2042-2049
• /
• 2019

Ten gizzard and three flesh samples of the broiler were collected from different locations in Tehsil Gujar Khan District Rawalpindi, Pakistan. The samples were dried, crushed and ground. Pellets were prepared by pressing the powder of the samples and that of the Bovine liver 1577c reference material obtained from NIST, USA. Proton induced X-ray emission (PIXE) installed at National Center for Physics, Islamabad, Pakistan has been used as a reliable and improved technique to determine concentration of various major/trace and toxic elements e.g. S, Cl, K, Ca, Cl, Fe, Cu, Mn, Co, Zn, Ti, Cd, Ga, Cr, V and Ni, in the Gizzard and Flesh samples of the broiler. The concentrations of all the detected elements in the samples are statistically significant. The certified and measured values of the elements in the reference material were in agreement with each other within a deviation of 7%. S, Cl, K and Ca are within tolerable limits and are good for human consumption. Ti, V, Cr, Mn, Fe, Co, Ni, Cu and Zn were more than the acceptable limits of World Health Organization, WHO whereas Ga, As, Sn, Sb and Pb are not detected in most of the samples.

• Journal of Nutrition and Health
• /
• v.20 no.2
• /
• pp.104-110
• /
• 1987

Thirty mael Sprague Dawley rats of 3 weeks of age were adopted to investigate changes in nutritive quality of the traditional, fermented soybean foods, Meju and Chungkugjang prepared by culturing with Aspergillus oryzae and the rice straw, respectively. The levels of each dietary protein were set at 10% . The effect of supplementation of fermented soybean on the rat diet was evaluated by measuring growth rate, feed efficiency. weight of the pancreas and biological values, such as NPU, PER and NPR on day 10. After incubation of soybean with Aspergilllus oryzae and the rice straw, the contents of crude protein, crude fat and crude ash were increased, but the levels of crude fiber and nitrogen-free extract were reduced. The amounts of feed intake, body weight gain an feed efficiency were greater for the rats fed dietscontaining heated unfermented soybeans or Meju compared to rats fed the diet containing raw soybeans. Pancreating enlargement was observed in the rats fed diets containing raw soybeans. The values of NPR and PER in rats fed diets containing Meju and Chungkugjang were significantly higher than those of rats fed diets containing raw soybeans. The values of NPU in rats fed diets containing raw soybans, heated unfermeted soybeans. Meju and Chungkugjang were 39.40%, 40.60%, 45.00% and 46.205, respectively, demonstrating no significant differences.