1.总结下这篇论文Thepapertitled"DeepPatchVisualOdometry(DPVO)"presentsanewdeeplearning-basedsystemformonocularvisualodometry(VO),whichestimatesarobot'spositionandorientationusingvideoinput.Keyhighlightsofthesystem

来自：https://github.com/princeton-vl/DPVO/blob/c0c5a104c9c58663aa9be62c3f125d5b52874f3e/dpvo/altcorr/correlation.py#L33classPatchLayer(torch.autograd.Function):@staticmethoddefforward(ctx,net,coords,radius):"""forwar