Efficient source camera identification with diversity-enhanced patch selection and deep residual prediction

Keyphrases

• Brand Model 25%
• Conventional Features 25%
• Convolutional Neural Network 25%
• Deep Learning 25%
• Deep Residual 100%
• Engineering Algorithm 25%
• Enhanced Diversity 25%
• Evaluation Protocol 25%
• Feature Engineering 25%
• Identification Method 25%
• Identification Performance 25%
• Image Content 25%
• Image Forensics 25%
• Image Patch 25%
• Instance-level 25%
• Level Evaluation 25%
• Multi-criteria 25%
• Patch Level 50%
• Patch Selection 100%
• Performance Comparison 25%
• Prediction Module 25%
• Reduced Training 25%
• Representative Patch 25%
• Residual Prediction 100%
• Scene Content 25%
• Small Images 25%
• Source Camera Identification 100%
• Training Costs 25%
• Training Data 25%
• VGGNet 25%

Engineering

• Convolutional Neural Network 100%
• Deep Learning 100%
• Engineering 100%
• Experimental Result 100%
• Image Content 100%
• Image Patch 100%
• Multiple Criterion 100%
• Multiscale 100%

Computer Science

• Convolutional Neural Network 50%
• Deep Learning 50%
• Experimental Result 50%
• Forensic Image 50%
• Image Content 50%
• Instance Level 50%
• Multiple Criterion 50%
• Performance Comparison 50%
• Prediction Residual 100%
• Training Data 50%