CRF with deep class embedding for large scale classification

Keyphrases

• Parametrization 100%
• Class Embedding 100%
• Large-scale Classification 100%
• Nonlinear Objective Function 50%
• Structured Objects 50%
• Learnability 50%
• Contextual Information 50%
• Visual Features 50%
• Structure Information 50%
• Class number 50%
• Highly Nonlinear 50%
• Deep Learning Architectures 50%
• Semantic Similarity 50%
• Image Sequence 50%
• Context-based 50%
• Class Information 50%
• Similarity Space 50%
• Visual Similarity 50%
• Class Structure 50%
• Retail Stores 50%
• Contextual Relations 50%
• Likelihood Maximization 50%
• Product Display 50%
• Pairwise Potential 50%
• Convolutional Layer 50%
• Potential Matrix 50%
• Huge Datasets 50%
• Linear-chain CRF 50%
• Batch Normalization 50%
• Ultrafine Grain 50%
• Likelihood Approximation 50%
• OCR Datasets 50%
• Surrogate Likelihood 50%
• Local Visual Features 50%
• Local Likelihood 50%

Computer Science

• Parametrization 100%
• Visual Feature 100%
• Approximation (Algorithm) 50%
• Recurrent Neural Network 50%
• Sequence Model 50%
• Contextual Information 50%
• Class Information 50%
• Optical Character Recognition 50%
• Learning Capacity 50%
• Batch Normalization 50%
• Class Structure 50%
• Convolutional Layer 50%
• Deep Learning 50%
• Objective Function 50%