We present a compact but effective CNN model for optical flow, calledPWC-Net. PWC-Net has been designed according to simple and well-establishedprinciples: pyramidal processing, warping, and the use of a cost volume. Castin a learnable feature pyramid, PWC-Net uses the cur- rent optical flowestimate to warp the CNN features of the second image. It then uses the warpedfeatures and features of the first image to construct a cost volume, which isprocessed by a CNN to estimate the optical flow. PWC-Net is 17 times smaller insize and easier to train than the recent FlowNet2 model. Moreover, itoutperforms all published optical flow methods on the MPI Sintel final pass andKITTI 2015 benchmarks, running at about 35 fps on Sintel resolution (1024x436)images. Our models are available on https://github.com/NVlabs/PWC-Net.