Ha, Mai Lan: Understanding images via visual similarity and deep feature representations. 2020
Inhalt
- Declaration of Authorship
- Zusammenfassungen
- Abstract
- Acknowledgements
- Contents
- List of Figures
- List of Tables
- Introduction
- Convolutional Neural Networks
- I Perceived Similarity
- Pixel-Based Perception-Inspired Metric for Intrinsic Imaging
- Perceptual Color Composition Similarity
- Related Work For Perceptual Color Similarity
- Hand-crafted Features for Color Similarity
- Learned Features for Visual Similarity
- Datasets for Perceptual Similarity
- Process to Define Perceptual Color Composition Similarity
- Computational Model of Color Composition Similarity
- Application 1: Global Color Descriptor
- Application 2: Fine-Grained Image Retrieval
- Related Work
- Content versus Color Retrieval
- Features and Training Model for Fine-Grained Retrieval
- Results Analysis and Discussion
- Analogy Image Retrieval - An Inspiration
- Application 3: Neural Style Transfer with Perceptual Color Similarity
- Related Work
- Perceptual Color Transfer
- Combining Style Transfer and Perceptual Color Transfer
- Results and Analysis
- More Style-Color Transfer Results
- More color transfer results
- Chapter Summary
- II Deep Features
- Shape Extraction and Semantic Segmentation
- Neural Discriminant Analysis
- Introduction To Discriminant Analysis For Fine-Grained Visual Classification
- Related Work
- Two-phase Neural Discriminant Analysis (NDA)
- Pre-optimized Features Extracted from Pre-trained DCNNs
- Feature Discriminant Analysis
- Discriminant Analysis Optimization with Neural Networks
- Experiments with Two-phase NDA
- Combined Optimization NDA
- Chapter Summary
- Conclusion
- Publications
- External Bibliography