EDNet: Efficient Disparity Estimation with Cost Volume Combination and Attention-based Spatial Residual

Keyphrases

• Attention-based 100%
• Disparity Estimation 100%
• Spatial Residuals 100%
• Cost Volume 100%
• Memory Consumption 66%
• 3D Convolutional Neural Network (3D CNN) 66%
• Similarity Measure 33%
• Neural Network 33%
• Contextual Information 33%
• Existing State 33%
• Residual Learning 33%
• Fast Speed 33%
• 3D Convolution 33%
• Scene Flow 33%
• Attention Mechanism 33%
• Feature Similarity 33%
• Learning Efficiency 33%
• Art Performance 33%
• Attention-aware 33%
• Inference Speed 33%
• Error Map 33%
• Correlation Volume 33%
• KITTI Dataset 33%
• Residual Module 33%
• Residual Features 33%
• 2D Convolution 33%

Computer Science

• Convolutional Neural Network 100%
• Memory Consumption 100%
• Similarity Measurement 50%
• Contextual Information 50%
• Attention (Machine Learning) 50%
• Art Performance 50%