Closed zhanghongyong123456 closed 1 year ago
Our method is based on low-resolution networks and are slower than pure low-resolution matting methods such as RVM, but much faster than patch-based inference. The inference time depends on the image resolution and the sparsity.