Learning the Distribution of Errors in Stereo Matching for Joint Disparity and Uncertainty Estimation
This work addresses the need for precise uncertainty estimates in stereo matching, which is incremental as it builds on multi-task learning approaches.
The paper tackled the problem of joint disparity and uncertainty estimation in stereo matching by introducing a new loss function that aligns uncertainty distribution with disparity errors using KL divergence, resulting in significant improvements in both tasks on large datasets.
We present a new loss function for joint disparity and uncertainty estimation in deep stereo matching. Our work is motivated by the need for precise uncertainty estimates and the observation that multi-task learning often leads to improved performance in all tasks. We show that this can be achieved by requiring the distribution of uncertainty to match the distribution of disparity errors via a KL divergence term in the network's loss function. A differentiable soft-histogramming technique is used to approximate the distributions so that they can be used in the loss. We experimentally assess the effectiveness of our approach and observe significant improvements in both disparity and uncertainty prediction on large datasets.