Please use this identifier to cite or link to this item: https://hdl.handle.net/10356/76946
Title: Scene text detection using neural network
Authors: Tan, Yi Hong
Keywords: DRNTU::Engineering::Computer science and engineering
Issue Date: 2019
Abstract: The scene text detection formula that uses region-based convolutional neural network (RCNN), is extremely standard in recent years. It boosts the performance considerably by creating a mix of 2 key insights. the primary one is to localize and section objects by applying high-capacity convolutional neural network to bottom-up region proposals. The second is to use supervised pre-training followed by domain-specific finetuning to coaching giant convolutional neural networks once the tagged training information is lean [1]. we tend to train a replacement model supported a dataset collected by ourselves from google web site. In this project, we tend to first of all introduce some relevant ideas. we tend to describe the selective search technique used for region proposal, and also the compositions of building blocks of convolutional neural network (CNN) that embrace convolutional layer, pooling layer, corrected linear units (ReLU), fully-connected layer and loss layer. we tend to detail the method of information set preparation that consists of data assortment, information labelling and format transformation, that is taken into account as vital preparation work for coaching method.The core part of this project is that the institution of the scene text detection system from the module style to the model testing and validation.
URI: http://hdl.handle.net/10356/76946
Rights: Nanyang Technological University
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:SCSE Student Reports (FYP/IA/PA/PI)

Files in This Item:
File Description SizeFormat 
Scene Text Detection Using Neural Network(Report).pdf
  Restricted Access
1.94 MBAdobe PDFView/Open

Page view(s)

101
Updated on Jun 22, 2021

Download(s) 50

47
Updated on Jun 22, 2021

Google ScholarTM

Check

Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.