Please enable JavaScript.
Coggle requires JavaScript to display documents.
Topics of deep learning for stereo (Ideas (Pre-processed (Gradients…
Topics of deep learning for stereo
Adversarial training
Recurrent process
Basics
Unsuppervised
Smooth loss
Left-right consistency
Censum transform
SSIM
Feature
What feature could we get from the noisy image
Ideas
Not only once.
Use the same CNN multiple times to get desired receptive field on different input image sizes?
Interact with the input.
Synthetic intermediate view.
Adaptive.
Uncalibrated LIDAR.
Normalized disparity.
Relative depth.
End-to-end disparity refinement.
Pre-processed
Feature points.
Gradients
Quantized