DCVSMNet: Double Cost Volume Stereo Matching Network
This is an incremental improvement for computer vision applications like 3D reconstruction, offering faster inference while maintaining accuracy.
The paper tackles stereo matching by introducing DCVSMNet, a novel architecture with two cost volumes and a coupling module, achieving competitive accuracy on benchmark datasets with a 67 ms inference time.
We introduce Double Cost Volume Stereo Matching Network(DCVSMNet) which is a novel architecture characterised by by two small upper (group-wise) and lower (norm correlation) cost volumes. Each cost volume is processed separately, and a coupling module is proposed to fuse the geometry information extracted from the upper and lower cost volumes. DCVSMNet is a fast stereo matching network with a 67 ms inference time and strong generalization ability which can produce competitive results compared to state-of-the-art methods. The results on several bench mark datasets show that DCVSMNet achieves better accuracy than methods such as CGI-Stereo and BGNet at the cost of greater inference time.