3D-RCNN: Instance-level 3D Object Reconstruction via Render-and-Compare
Wepresentafastinverse-graphicsframeworkforinstance-level3Dsceneunderstanding.Wetrainadeepconvolutionalnetworkthatlearnstomapimageregionstothefull3Dshapeandposeofallobjectinstancesintheimage.Ourmethodproducesacompact3Drepresentationofthescene,whichcanbereadilyusedforapplicationsli
用户评论