PSMNet is a pyramid stereo matching network that can be used for depth estimation. It consists of two main modules: spatial pyramid pooling and 3D CNN. The spatial pyramid pooling module takes advantage of the capacity of global context information by aggregating context in different scales and locations to form a cost volume. The 3D CNN regularizes cost volume using stacked multiple hourglass networks with intermediate supervision.
The following image shows the result of PSMNet.
The following table lists the PSMNet models supported by the Vitis AI Library.