NVlabs / neuralrgbd

Neural RGB→D Sensing: Per-pixel depth and its uncertainty estimation from a monocular RGB video
Other
301 stars 41 forks source link

question about DPV estimation of D-Net #20

Closed Yvaine closed 3 years ago

Yvaine commented 3 years ago

Hi, I have one question. In estimate DPV of D-Net, after calculate the residual(cost volume) of feature of two frames, when the residual is small, that represent the two frames feature is similar, because the probability of depth is the inverse of softmax of residual, so, my question is why similar feature represent the probability of depth is big? What's the relation between feature and probability of depth? Thank you.