Portrait matting using an attention-based memory network

Keyphrases

• Matting 100%
• Attention-based 100%
• Memory Network 100%
• Portrait Matting 100%
• Self-attention 50%
• Temporal Coherence 50%
• Video Frames 50%
• Prediction Quality 50%
• Pixel Level 50%
• Memory Block 50%
• Quality Consistency 50%
• Network Robustness 50%
• Temporal Consistency 50%
• Network Attention 50%
• Trimap 50%
• Segmentation Mask 50%
• Auxiliary Input 50%
• Binary Segmentation 50%
• Direct Supervision 50%
• Foreground Estimation 50%
• Alpha Forecasting 50%
• Video Matting 50%
• Memory Module 50%