A Conditional Adversarial Network for Scene Flow Estimation
This work addresses computational overhead issues in scene flow estimation for robotics applications, representing an incremental advancement by applying GANs to this task for the first time.
The paper tackles the problem of scene flow estimation in depth videos by proposing SceneFlowGAN, a conditional adversarial network that simultaneously estimates optical flow and disparity from stereo images, achieving results on a large RGB-D benchmark dataset.
The problem of Scene flow estimation in depth videos has been attracting attention of researchers of robot vision, due to its potential application in various areas of robotics. The conventional scene flow methods are difficult to use in reallife applications due to their long computational overhead. We propose a conditional adversarial network SceneFlowGAN for scene flow estimation. The proposed SceneFlowGAN uses loss function at two ends: both generator and descriptor ends. The proposed network is the first attempt to estimate scene flow using generative adversarial networks, and is able to estimate both the optical flow and disparity from the input stereo images simultaneously. The proposed method is experimented on a large RGB-D benchmark sceneflow dataset.