An AI Approach Towards Civil Engineering Design Annotations and Object Detection

An AI Approach Towards Civil Engineering Design Annotations and Object Detection

Harshitha N1, Pragathi S2, Dr S Padmashree3

Department of ECE, GSSSIETW,MYSURU

ABSTRACT- Design is one of the principal steps in the creation challenge. Error in layout causes failure in the construction level as properly as inside the assignment improvement degree. Misunderstandings of design standards between parties lead to design mistakes which may additionally lower the first-rate of construction works. It also reasons cost overruns and venture delays.AI strategies can assist inexperienced device operators to hold out complex tasks, which machine operators otherwise couldnt. It is high-quality in the ones industries which might be locating it increasingly tough to source fairly skilled and skilled operators. The most important applications for AI inside the production sector appear to coagulate around via using gadget getting to know for pattern recognition as nicely as image reputation By the use of AI most of the huge creation companies have started out to advantage greater prominence because of its enhancing safety of human workers in creation sites..Hence the aim of our paper is to create a ML model capable of identifying errors and correcting errors automatically to reduce processing time and training model that learns from the existing data , drawing to recognize the error in the new assigned data and intimidate the errors to be viewed and corrected.

1. INTRODUCTION

   Artificial Intelligence is also called as machine intelligence, that's demonstrated with the aid of machines, in assessment to natural intelligence as displayed in humans. AI techniques have turn out to be an essential part of the technological industry, that's indeed supporting to solve diverse challenging troubles within the discipline of pc science, research, software engineering, and different fields The education set which has been given for supervised mastering is labeled dataset. Supervised getting to know attempts to discover the correlation between the feature set and the label set, that is the expertise and houses that we examine from the categorized dataset. The information acquired from supervised getting to know is regularly utilized for prediction and popularity. Unsupervised getting to know goals at clustering, dimensionality reduction, finding association among features, and opportunity density estimation. In general, an unmanaged algorithm may at the equal time study more than one property. The results of unsupervised mastering may be additionally used for supervised learning. Civil engineering is a expert engineering subject that offers with the design, construction, and renovation of the physical and evidently built environment, which include public works along with roads, airports, pipelines, dams, bridges, sewerage systems, canals, railways, and structural additives of buildings, Civil designs were evolving and growing extra

   time now. With the rapid development in this field, the structures demanding complex designing and construction, the verification of these designs takes more time and multilayers of reviewing which consumes more time and more engineers involved in the process which in turn decreases the efficiency ie in the time where engineers can be reviewing many drawings they just focus and give more time on just a few drawings.

   In this current work, an attempt is made to segment the civil engineering drawings through suitable segmentation algorithm for proper partitioning of the images and further through AI model annotation and correction of civil engineering terms are performed.

   The intelligent engineering design correction is a boon for the civil design industry since it reduces the human efforts and it saves time by reviewing the designs and identifying the faults and errors in the designs and predict the pattern of errors and corrections and corrects the deviations /errors, making it much easier and faster for the reviewing and correction process to complete.

2. LITERATURE SURVEY

   U. Karthikeyan and Dr. M. Vanitha proposed, A Study on Text Recognition the use of Image Processing with Data mining Techniques the use of Image segmentation and Edge detection, which presented the overview on distinctive aspect detection techniques used for picture segmentation [1]. Andreas Nürnberger, Christian Hentschel, Sebastian Stober, and Marcin Detynieck proposed a examine on, Automatic Image Annotation Using a Visual Dictionary Based on Reliable Image Segmentation, the use of Automatic Image Annotation (AIA) technique. This paper offered a framework to guard photograph annotations primarily based on a visual dictionary that's created automatically the use of a fixed of domestically annotated schooling images [2]. Annamma Abraham and R.Sathya proposed a take a look at on, Comparison of Supervised and Unsupervised Learning Algorithms for Pattern Classification. This paper examined the two gaining knowledge of algorithms particularly supervised and unsupervised, additionally they investigated its properties in the classification [3]. Mahima Singh Deo, S. Venkatesan, Annina Simon, and D.R.Ramesh Babu presented An assessment of gadget learning and its programs, which incorporates the Study of machine mastering

   An overview of machine learning and its applications, which includes the Study of machine learning and its applications [4]. Patrick Ozer proposed a study on Data Mining Algorithms for Classification, which includes the Study as well as comparison of supervised and unsupervised learning methods. This paper examined the two learning algorithms namely supervised and unsupervised. They also investigated its properties in the classification [5]. SehlaZayenz, Nadia Jmour, and Afef Abdelkrim propound a study on Convolutional neural networks for image classification. This paper narrated a learning approach based on the training convolutional neural networks (CNN) for the traffic sign classification system [6]. Ali Farhadi and Joseph Redmon submitted a study on YOLOv3: An Incremental Improvement. They have done a bunch of little changes in the design to make it better. They also trained this new network that is quite swelling. It is a little bigger than the last but faultless. Though it is fast, still YOLOv3 is a good detector

3. OVERVIEW OF LITERATURE SURVEY

   From all the above literature review it is inferred that The application and importance of machine learning in the nearing future and how it is going to reduce human interaction ,Overview on machine learning, supervised and unsupervised method of learning and its properties in machine learning and An overview of annotation and how locally annotated set of images can be used to train the model using the visual dictionary and also an incremental improvement made in YOLOv3 that recognizes 80 different objects in images,but most importantly it is fast and accurate.

4. METHODOLOGY

   STEP 1: Annotate images

   For our detector to learn to detect objects in images, it as to be fed with labeled training data.

   STEP 2: Training of Model

   Neural network structure: Image pixels are used for object detection as features while training a network or while making a prediction. If we use fully connected layers, where every neuron is connected to all the neurons in the previous layer then. Due to the huge number of pixels in an image, there is an increase in the weights of the layers. This leads to a better solution which can give us desired results.

   Fig 1:Three layered neural network with two fully connected layer

   Fig 2:Machine learning process

   Convolutional Neural Networks (CNN) is the type of neural network that consists of neurons, which have weights, biases, activation functions. Convolutional neural networks is a multilayer perceptron, which applies convolution functions from computer vision. Convolutional layers try to extract useful features from images and the neurons try to convert input signals into the appropriate output signals based on the automated reaction.

   Neural network training: The process loss refers to the error of the model and it shows how well the model is improving during the training. If the loss is decreased, the model training is going well, if it is constant it means that the model is already trained for set parameters. Data is provided to the algorithm in the form of batches, where a batch can possess multiple training samples. Training a single batch is considered as one iteration cycle, going over the whole dataset once is called an epoch. The model architecture which we are using is called YOLOv3 (You Only Look Once). This particular model is a one-shot learner, which means a particular image can only pass through the network once to make a prediction. This allows the architecture to be high performance, which can view up to 60 frames per second. YOLO divides an image into subcomponents and before pooling back to create a prediction it conducts convolutions on each of those subcomponents.

   STEP 3: Testing of the model

   Once our model is trained, it will be used to make predictions. For the purpose of prediction, we require the following

   (a) Setting up the YOLOv3 model architecture. (b)We train the architecture using the custom weights.

5. RESULT

   Model is trained efficiently and effectively using many database references available, which in turn gives the system the confidence and ability to identify the error or the portion to be corrected and to intimidate the engineer about the error and correct it on its own.

6. FUTURE SCOPE

Model in future can identify Errors and insights that human eyes cant notice without putting in a lot of effort. Moreover, it is faster and more scalable at doing so.

ACKNOWLEDGEMENT

This work is supported in part by SANRIA Engineering and consulting Pvt Ltd. A technology-driven engineering company which provides multidisciplinary solutions in connection to design, steel detailing, 3D modeling, As – Built Engineering, Building designed project Management consulting.

REFERENCES

1. U.Karthikeyan, Dr. M. Vanitha A Study on Text Recognition using Image Processing with Datamining Techniques , IJCSE All Rights Reserved 2019

2. Christian Hentschel, Sebastian Stober, Andreas Nürnberger, and Marcin Detyniecki ,Automatic;Image Annotation Using a Visual Dictionary Based on Reliable Image Segmentation,Otto-von- Guericke-University, Magdeburg Laboratoire dInformatique de Paris.

3. R. Sathya , Annamma Abraham Algorithms for Pattern Classification (IJARAI) International Journal of Advanced Research in Artificial Intelligence, Vol. 2, No. 2, 2013

4. Annina Simon,Mahima Singh Deo ,Venkatesan Selvam ,Ramesh Babu :An Overview of Machine Learning and its Applications International Journal of Advanced Research in Artificial Intelligence, Vol. 1,No1, 2015.

5. Patrick Ozer,Data Mining Algorithms for Classification,Radboud University Nijmegen January 2012.

6. Nadia Jmour ; Sehla Zayen ; Afef Abdelkrim ;Convolutional neural networks for image classification ; International Conference on Advanced Systems and Electric Technologies 2018

7. Joseph Redmon, Ali Farhadi,YOLOv3:An Incremental Improvement,University of Washington.

8. LI Huanliang ,The Research of Intelligent Image Recognition Technology Based on Neural Network .a Yishui Campus of Linyi University, Yishui 276400,China.

9. Nada Farhani; Naim Terbeh, Image to Text Conversion: State of the Art and Extended Work Publisher: IEEE; Mounir Zriguior.

10. Mohit Agarwal ; Baijnath Kaushik,Text recognition from image using artificial neural network and genetic algorithm.Publisher: IEEE .

11. Osisanwo F.Y.Akinsola J.E.T Awodele , Hinmikaiye J. Olakanmi Akinjobi J. Supervised Machine Learning Algorithms ,Classification and Comparison,International Journal of Computer Trends and Technology (IJCTT) Volume 48 No9, 2017.

12. H. Joel Trussell Eli Saber Michael Vrhel.Color image processing: Basics and special issue overview ,IEEE Signal processing magazine,

    Vol. 2 ,2015

13. Akhil S. Deshpande, Shreyas S. Vaidya, Pravin B. Swami, Pavan R. Jaiswal, Integration of Speech, Image & Text Processing Technologies ; International Journal of Advanced Research in Artificial Intelligence

14. Sumit Das, Aritra Dey, Akash ,Nabamita Roy;Applications of Artificial Intelligence in Machine Learning: Review and Prospect ; International Journal of Computer Applications Volume 115 No. 9 2015.

15. A Kajaree Das1 , Rabi Narayan Behera; Survey on Machine Learning: Concept, Algorithms and Applications;International Journal of Innovative Research in Computer and Communication Engineering 2017.

16. Taiwo Oladipupo Ayodele University of Portsmouth United Kingdom

17. Mohit Agarwal, Baijnath Kaushik; Text recognition from image using artificial neural network and genetic algorithm ;International Conference on Green Computing and Internet of Things (ICGCIoT) 2015

18. Dhanush Kumar S, Lavanya S ,Madhumita G ,Mercy Rajaselvi V; Speech to Text Conversion International Journal of Advance Research, Ideas and Innovations in Technology Volume 4 2018

19. Tatiana Stefanova ,Reporting in Tekla Structures .Bachelors thesis Construction Engineering Visamäki, Spring 2015 Tatiana Stefanova .

20. Rupneet Kaur Hanspal1 , Kishor Sahoo; A Survey of Image Enhancement Techniques;International Journal of Science and Research (IJSR) Volume 6 Issue 5, May2017

21. Shubhashree Savant ;A Review on Edge Detection Techniques for Image Segmentation ;(IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 5 (4) , 2014

22. Marcia A Bush, Speech And Text Image Processing In Documents, Xerox Palo Alto Research Center 3333 Coyote HILL Road Palo Alto CA94304.