Object Detection and Recognition Techniques for Indoor Scene

发布时间：2020-12-26 16:42

　　Object detection is one of the most classical computer vision task that is used to detect objects from an image.Efficient and accurate object detection plays a vital role in the advancement of computer vision that makes it able to detect the objects in real time applications with more accuracy and less processing time.With recent advancement in machine learning and deep learning techniques,the accuracy for object detection has significantly increased.Existing Convolutional Neural Network（CNN）bas... 

【文章来源】：华南理工大学广东省 211工程院校 985工程院校 教育部直属院校

【文章页数】：93 页

【学位级别】：硕士

【文章目录】：
Abstract
Abbreviations and Acronyms
1 Chapter 1: Introduction
    1.1 Introduction
    1.2 Problem Statement
    1.3 Thesis Structure
2 Chapter 2: Overview of CNN
    2.1 Neural Networks
        2.1.1 Neural Network Background
        2.1.2 Multi-Layer Networks
        2.1.3 Back-Propagation
        2.1.4 Activation function
    2.2 Deep Learning
    2.3 Computer Vision
        2.3.1 Object Detection
    2.4 Convolutional Neural Networks
        2.4.1 Working principle of Convolutional Neural Network
        2.4.2 Pooling and Stride
        2.4.3 Layers in Convolutional Neural Network
        2.4.4 Regularization and Data Augmentation
    2.5 CNN Architecture
        2.5.1 AlexNet
        2.5.2 ZFNet
        2.5.3 VGGNet
        2.5.4 GoogLeNet/Inception
        2.5.5 ResNet
3 Chapter 3: Object Detection and Classification
    3.1 Object Detection
    3.2 Object Detection Techniques
        3.2.1 SIFT Detector
            3.2.1.1 PCA-SIFT Detector
            3.2.1.2 SURF Detector
        3.2.2 Bag-of-visual-words
    3.3 CNN Based Techniques
        3.3.1 R-CNN
            3.3.1.1 R-CNN Structure
            3.3.1.2 Drawbacks
        3.3.2 Fast R-CNN
            3.3.2.1 Classification and Performance
            3.3.2.2 Training
    3.4 Region Proposal Based Object Detection Technique
        3.4.1 Overview
            3.4.1.1 Dense Set Generation
            3.4.1.2 Sparse Set Generation
        3.4.2 Selective Search Parameters
        3.4.3 Edge Box
    3.5 Faster R-CNN
    3.6 SSD
    3.7 Comparison of CNN Based Object Detection Techniques
    3.8 Standard Benchmarks
    3.9 Categorize of Object Detection
        3.9.1 Objectness Detection
            3.9.1.1 Window-sliding Method
            3.9.1.2 Region-merging Method
            3.9.1.3 Box-regressing Method
        3.9.2 Salient object detection
            3.9.2.1 Bottom-up SOD
            3.9.2.2 Top-down SOD:
        3.9.3 Category Specific Object Detection
            3.9.3.1 Object proposal-based Methods
            3.9.3.2 Regression-based Methods
4 Chapter 4: Proposed Techniques
    4.1 Proposed Techniques
    4.2 Indoor-V Dataset
    4.3 Indoor-VI Dataset
        4.3.1 Resizing
        4.3.2 Rotation
        4.3.3 Flipping
        4.3.4 Lighting conditions
        4.3.5 Noise adding
        4.3.6 Blurring
    4.4 Image Annotation
    4.5 Converting of XML to CSV and CSV to TF Record
    4.6 System Implementation and Experiments
    4.7 Proposed Technique 1 Architecture
        4.7.1 Parameters
        4.7.2 Fine-Tuning the Network
        4.7.3 Training 1st Proposed Technique
        4.7.4 Evaluation of 1st Proposed Technique
        4.7.5 Results Analysis of Proposed Technique 1
            4.7.5.1 Results of Proposed Technique 1 for Indoor-V Dataset
            4.7.5.2 Results of Proposed Technique 1 For Indoor-VI Dataset
    4.8 Proposed Technique 2 Architecture
        4.8.1 Parameters
        4.8.2 Fine-Tuning the Network
        4.8.3 Training 2nd Proposed Technique
        4.8.4 Evaluation 2nd Proposed Technique
        4.8.5 Results Analysis of Proposed Technique 2
            4.8.5.1 Results of Proposed Model 2 on Indoor-V Dataset
            4.8.5.2 Results of Proposed Technique 2 for Indoor-VI Dataset
    4.9 Comparison with Existing Techniques
5 Chapter 5: Conclusion
    5.1 Conclusion
    5.2 Future work
6 BIBLIOGRAPHY
7 Appendix
    7.1 Proposed Technique 1
    7.2 Proposed Technique 2
8 Acknowledgements
附表


【参考文献】：
期刊论文
[1]数据大爆炸是我们时代的创新故事[J]. Erik Brynjolfsson,Andrew McAfee,刘白云.  英语文摘. 2012(02)



本文编号：2940076