• 제목/요약/키워드: pruning method

검색결과 170건 처리시간 0.032초

A Lightweight Pedestrian Intrusion Detection and Warning Method for Intelligent Traffic Security

  • Yan, Xinyun;He, Zhengran;Huang, Youxiang;Xu, Xiaohu;Wang, Jie;Zhou, Xiaofeng;Wang, Chishe;Lu, Zhiyi
    • KSII Transactions on Internet and Information Systems (TIIS)
    • /
    • 제16권12호
    • /
    • pp.3904-3922
    • /
    • 2022
  • As a research hotspot, pedestrian detection has a wide range of applications in the field of computer vision in recent years. However, current pedestrian detection methods have problems such as insufficient detection accuracy and large models that are not suitable for large-scale deployment. In view of these problems mentioned above, a lightweight pedestrian detection and early warning method using a new model called you only look once (Yolov5) is proposed in this paper, which utilizing advantages of Yolov5s model to achieve accurate and fast pedestrian recognition. In addition, this paper also optimizes the loss function of the batch normalization (BN) layer. After sparsification, pruning and fine-tuning, got a lot of optimization, the size of the model on the edge of the computing power is lower equipment can be deployed. Finally, from the experimental data presented in this paper, under the training of the road pedestrian dataset that we collected and processed independently, the Yolov5s model has certain advantages in terms of precision and other indicators compared with traditional single shot multiBox detector (SSD) model and fast region-convolutional neural network (Fast R-CNN) model. After pruning and lightweight, the size of training model is greatly reduced without a significant reduction in accuracy, and the final precision reaches 87%, while the model size is reduced to 7,723 KB.

사진과 전문가 인터뷰를 통해 추론한 근대 궁궐의 수목관리 판단 연구 (A Study on the Types of Tree Management in Modern Palace Using Photographs and Expert Interviews)

  • 최진서;김충식
    • 한국조경학회지
    • /
    • 제51권2호
    • /
    • pp.94-102
    • /
    • 2023
  • 1981년 ICOMOS-IFLA 국제 역사 정원 위원회에서 제정한 플로렌스 헌장에서는 역사정원을 식물이 주를 이루는 건축적 구성으로 이를 영원히 변하지 않도록 유지하고자 하는 예술가와 장인의 욕구사이에 존재하는 끊임없는 균형으로 보았다. 이처럼 정원의 주된 구성요소인 수목은 계절의 순환에 따라 생성과 소명을 반복하기 때문에 지속적인 관리가 필요하다. 이에 따라 우리나라 궁궐에서도 수목의 모습을 유지하기 위한 관리는 필수불가결한 요소였을 것이다. 다만 과거 궁궐의 수목관리 기법을 고증하는 것은 매우 중요한 일이지만 역사적 기록 부재와 일제강점기로 인해 명맥이 단절됨에 따라 연구가 어려운 실정이다. 또한 일반적으로 궁궐의 수목은 관리를 하지 않았다는 견해에 따라 궁궐 수목 관리 기법에 관한 연구는 지금까지 수행되지 않았다. 본 연구는 근대에 촬영된 사진을 토대로 전문가 인터뷰를 통해 과거 궁궐의 수목관리 판단 여부를 밝히는데 목적을 두었다. 근대기에 촬영된 사진을 활용하여 전문가에게 심층 인터뷰를 통해 수종의 식별과 전정여부를 파악한 내용을 토대로 결과는 다음과 같다. 첫째, 사진을 통해 수목의 수형과 잎 형태 식별이 가능함을 확인하였으며 정지·전정 등에 의해서 생기는 현상을 관찰함으로써 근대기 궁궐의 수목관리 시행 여부를 추정할 수 있었다. 둘째, 4개 분야에서 8명의 전문가들에게 심층 인터뷰를 한 결과 수종 식별, 전정여부 및 목적, 방법 등의 관리여부의 판별이 가능하였으며 집단별 의견의 차이가 크게 발생하지 않고 근거를 명확하게 제시하였다. 셋째, 궁궐 수목의 관리 유형은 수형관리, 수목의 위해 요인 제거, 하층식생관리가 주된 것으로 판단하였으며 존덕정, 관람정 등의 사진을 통해 촬영 시점인 일제강점기 이전에도 수목의 관리가 이루어졌음을 확인하였다. 촬영된 사진을 토대로 전문가 인터뷰를 거쳐 일제강점기 이전 수목관리 여부 추정이 가능하였다. 그러나 당시 시대 상황에 따라 자체적으로 수행된 것인지 일제에 의해 수행된 것인지는 사료의 부족으로 규명하지 못하였다. 하지만, 과거 궁궐의 수목관리를 하지 않았다는 견해를 수집된 자료를 통해 반박할 수 있는 근거자료를 마련하였으며 이를 뒷받침하는 전문가 의견을 종합하여 여부를 판단하였다. 또한 일반적인 정지·전정 이론을 토대로 전문가 의견에 대한 실증적인 검토를 실시하여 연구결과에 신뢰성을 확보하였다.

소셜 미디어 상에서 개인화된 여행 경로 추천 기법 (Personalized Travel Path Recommendation Scheme on Social Media)

  • ;임종태;복경수;유재수
    • 한국콘텐츠학회논문지
    • /
    • 제19권2호
    • /
    • pp.284-295
    • /
    • 2019
  • 소셜 미디어 환경에서 여행과 커뮤니티에서 기고한 사진과 관련된 메타 데이터 (태그, 지리적 위치 및 찍은 날짜)에 기반한 개인화 된 여행 경로 추천 기법이 연구되고 있다. 사용자는 소설 미디어를 사용하고 자신의 위치 기록을 여행 경로의 형태로 기록한다. 이러한 여행 경로 정보는 미래의 여행자들에게 새로운 추천 정보를 제공하기 위한 유용한 정보로 활용 될 수 있다. 본 논문에서는 라이프 로그를 기반으로 한 개인화 된 여행 경로 추천 기법을 제안한다. 제안하는 기법은 여행자 및 지역 사회가 제공한 라이프 로그 및 사진 정보를 활용하여 사용자에게 개인화된 추천 서비스를 제공할 수 있을 뿐만 아니라 개별 관심 장소가 아닌 대중적인 여행 경로도 추천 할 수 있다 (POI). 제안하는 개인화된 여행 경로 추천 기법은 POI 가지치기 단계와 여행 경로 생성 단계로 구성된다. POI 가지치기 단계에서는 POI 전체 데이터로부터 사용자에게 추천할 경로를 생성하는데 필요한 POI만을 남기고 가치기를 수행한다. 여행 경로 생성 단계에서는 POI 가지치기 단계를 통해 도출된 POI 사용자 관심도, 비용, 시간, 이벤트 등을 고려하여 후보 경로를 생성한다.

하한비용 추정에 바탕을 둔 최적 스케쥴링기법 (An Optimal Scheduling Method based upon the Lower Bound Cost Estimation)

  • 엄성용;전주식
    • 전자공학회논문지A
    • /
    • 제28A권12호
    • /
    • pp.73-87
    • /
    • 1991
  • This paper presents a new approach to the scheduling problem in the high level synthesis. In this approach, iterative rescheduling processes starting with ASAP(As Soon As Possible) scheduling result are performed in a branch-and-bound manner so to arrive at the scheduling result of the lowest hardware cost under the given timing constraint. At each iteration step, only the selected nodes are considered for rescheduling, and the lower bound cost estimation is performed to avoid the unnecessary attempts to search for an optimal result. This branch-and-bound method turns out to be effective in pruning the search space, and thus reducing run time considerably in many cases.

  • PDF

BAYESIAN CLASSIFICATION AND FREQUENT PATTERN MINING FOR APPLYING INTRUSION DETECTION

  • Lee, Heon-Gyu;Noh, Ki-Yong;Ryu, Keun-Ho
    • 대한원격탐사학회:학술대회논문집
    • /
    • 대한원격탐사학회 2005년도 Proceedings of ISRS 2005
    • /
    • pp.713-716
    • /
    • 2005
  • In this paper, in order to identify and recognize attack patterns, we propose a Bayesian classification using frequent patterns. In theory, Bayesian classifiers guarantee the minimum error rate compared to all other classifiers. However, in practice this is not always the case owing to inaccuracies in the unrealistic assumption{ class conditional independence) made for its use. Our method addresses the problem of attribute dependence by discovering frequent patterns. It generates frequent patterns using an efficient FP-growth approach. Since the volume of patterns produced can be large, we propose a pruning technique for selection only interesting patterns. Also, this method estimates the probability of a new case using different product approximations, where each product approximation assumes different independence of the attributes. Our experiments show that the proposed classifier achieves higher accuracy and is more efficient than other classifiers.

  • PDF

Neural Networks Based Modeling with Adaptive Selection of Hidden Layer's Node for Path Loss Model

  • Kang, Chang Ho;Cho, Seong Yun
    • Journal of Positioning, Navigation, and Timing
    • /
    • 제8권4호
    • /
    • pp.193-200
    • /
    • 2019
  • The auto-encoder network which is a good candidate to handle the modeling of the signal strength attenuation is designed for denoising and compensating the distortion of the received data. It provides a non-linear mapping function by iteratively learning the encoder and the decoder. The encoder is the non-linear mapping function, and the decoder demands accurate data reconstruction from the representation generated by the encoder. In addition, the adaptive network width which supports the automatic generation of new hidden nodes and pruning of inconsequential nodes is also implemented in the proposed algorithm for increasing the efficiency of the algorithm. Simulation results show that the proposed method can improve the neural network training surface to achieve the highest possible accuracy of the signal modeling compared with the conventional modeling method.

Extracting the K-most Critical Paths in Multi-corner Multi-mode for Fast Static Timing Analysis

  • Oh, Deok-Keun;Jin, Myeoung-Woo;Kim, Ju-Ho
    • JSTS:Journal of Semiconductor Technology and Science
    • /
    • 제16권6호
    • /
    • pp.771-780
    • /
    • 2016
  • Detecting a set of longest paths is one of the crucial steps in static timing analysis and optimization. Recently, the process variation during manufacturing affects performance of the circuit design due to nanometer feature size. Measuring the performance of a circuit prior to its fabrication requires a considerable amount of computation time because it requires multi-corner and multi-mode analysis with process variations. An efficient algorithm of detecting the K-most critical paths in multi-corner multi-mode static timing analysis (MCMM STA) is proposed in this paper. The ISCAS'85 benchmark suite using a 32 nm technology is applied to verify the proposed method. The proposed K-most critical paths detection method reduces about 25% of computation time on average.

근접방어무기체계에 대한 함대함 유도탄의 최적회피기동 (Optimal Evasive Maneuver for Sea Skimming Missiles against Close-In Weapon System)

  • 황익호
    • 대한전기학회:학술대회논문집
    • /
    • 대한전기학회 2002년도 하계학술대회 논문집 D
    • /
    • pp.2096-2098
    • /
    • 2002
  • In this paper, the optimal evasive maneuver strategies for typical subsonic ASM(anti-ship missile) to reach its target ship with high survivability against CIWS(close in weapon system) are studied. The optimal evasive maneuver input is defined by the homing command optimizing the cost function which takes aiming errors of CIWS into account. The optimization problem for the effective evasive maneuver is formulated based on a simple missile dynamics model and a CIWS model. By means of solving the problem, a multiple hypotheses testing method is proposed. Since this method requires generation of too many hypotheses, the hypothesis-pruning technique is adopted. The solution shows that the optimal evasive maneuver is a bang-bane shaped command whose frequency is varied by the aimpoint determination strategy in CIWS.

  • PDF

Optimized Neurocontroller for Human Control Skill Transfer

  • Seo, Kap-Ho;Changmok Oh;Lee, Ju-Jang
    • 제어로봇시스템학회:학술대회논문집
    • /
    • 제어로봇시스템학회 2001년도 ICCAS
    • /
    • pp.42.3-42
    • /
    • 2001
  • A human is an expert in manipulation. We have acquired skills to perform dexterous operations based upon knowledge and experience attained over a long period of time. It is important in robotics to understand these human skills, and utilize them to bring about better robot control and operation It is hoped that the neurocontroller can be trained and organized by simply presenting human teaching data, which implicate human intention, strategy and expertise. In designing a neurocontroller, we must determine the size of neurocontroller. Improper size may not only incur difficulties in training neural nets, e.g. no convergence, but also cause instability and erratic behavior in machines. Therefore, it is necessary to determine the proper size of neurocontroller for human control transfer. In this paper, a new pruning method is developed, based on the penalty-term methods. This method makes ...

  • PDF

Transmission of Apple scar skin viroid by Grafting, Using Contaminated Pruning Equipment, and Planting Infected Seeds

  • Kim, Hyun-Ran;Lee, Sin-Ho;Lee, Dong-Hyuk;Kim, Jeong-Soo;Park, Jin-Woo
    • The Plant Pathology Journal
    • /
    • 제22권1호
    • /
    • pp.63-67
    • /
    • 2006
  • Apple scar skin, one of the most destructive diseases affecting apple, is caused by Apple scar skin viroid (ASSV d). Fruit dappling appeared on several cultivars in Korea and has been distributed to major cultivated areas since 2001. ASSVd was identified from infected fruits by using nucleic acid sequence-based amplification with electrochemiluminescence (NASBA-ECL). NASBA-ECL method was faster and hundredfold more sensitive than reverse transcription-polymerase chain reaction (RT-PCR) for ASSVd detection in apple leaves/ stems. ASSVd was rapidly transmitted to the entire tree in the second year after artificial inoculation. The ASSVd could be transmitted efficiently by using contaminated pruning scissors to both lignified stems (60 to $70\%$) and green shoots (20 to $40\%$) of apple tree and young plants. Dipping of contaminated scissors in $2\%$ sodium hypochlorite solution effectively prevented viroid transmission. In the ASSV d-infected fruits, the viroid was easily detected from fruit skin, seed coat, and embryo. Moreover, embryo and endosperm separately excised from the ASSVd-infected seeds were ASSVd positive in NASBA-ECL assay. Seedlings germinated from ASSVd-positive seeds showed $7.7\%$ infection rate., which indicated that ASSVd is seed-borne.