A Multi-instance Multi-label Dual Learning Approach for Video Captioning

Keyphrases

• Benchmark Dataset 20%
• Caption Generation 60%
• Challenging Tasks 20%
• Dual Learning 100%
• Encoder-decoder 40%
• Encoder-decoder Structure 20%
• Fully Convolutional Neural Network 20%
• Generation Module 40%
• Learning Approaches 100%
• Learning Mechanisms 40%
• Multi-label Learning 100%
• Multimedia Processing 20%
• Natural Language Description 20%
• Raw Video 40%
• Reconstruction Module 40%
• Reconstructor 40%
• Semantic Gap 20%
• Two-module 40%
• Unsupervised Learning 20%
• Video Caption 60%
• Video Captioning 100%
• Video Content 20%
• Video Reconstruction 40%
• Visual Information 20%
• Visual Sequences 40%

Computer Science

• Convolutional Neural Network 25%
• Experimental Result 25%
• Learning Approach 100%
• Learning Mechanism 50%
• Multimedia 25%
• semantic gap 25%
• Video Reconstruction 50%
• Visual Informations 25%