?url_ver=Z39.88-2004&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Adc&rft.title=DynamicSurf%3A+Dynamic+Neural+RGB-D+Surface+Reconstruction+With+an+Optimizable+Feature+Grid&rft.creator=Mohamed%2C+M&rft.creator=Agapito%2C+L&rft.description=We+propose+DynamicSurf%2C+a+model-free+neural+implicit+surface+reconstruction+method+for+high-fidelity+3D+modelling+of+non-rigid+surfaces+from+monocular+RGB-D+video.+To+cope+with+the+lack+of+multi-view+cues+in+monocular+sequences+of+deforming+surfaces%2C+one+of+the+most+challenging+settings+for+3D+reconstruction%2C+DynamicSurf+exploits+depth%2C+surface+normals%2C+and+RGB+losses+to+improve+reconstruction+fidelity+and+optimisation+time.+DynamicSurf+learns+a+neural+deformation+field+that+maps+a+canonical+representation+of+the+surface+geometry+to+the+current+frame.+We+depart+from+current+neural+non-rigid+surface+reconstruction+models+by+designing+the+canonical+representation+as+a+learned+feature+grid+which+leads+to+faster+and+more+accurate+surface+reconstruction+than+competing+approaches+that+use+a+single+MLP.+We+demonstrate+DynamicSurf+on+public+datasets+and+show+that+it+can+optimize+sequences+of+varying+frames+with+6%C3%97+speedup+over+pure+MLP-based+approaches+while+achieving+comparable+results+to+the+state-of-the-art+methods.&rft.subject=Geometry%2C+Surface+reconstruction%2C+Solid+modeling%2C%0D%0AThree-dimensional+displays%2C+Deformation%2CNetwork+topology%2C%0D%0ADynamics&rft.publisher=IEEE&rft.date=2024-06-12&rft.type=Proceedings+paper&rft.language=eng&rft.source=+++++In%3A++2024+International+Conference+on+3D+Vision+(3DV).++(pp.+pp.+820-830).++IEEE%3A+Davos%2C+Switzerland.+(2024)+++++&rft.format=application%2Fpdf&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10194974%2F1%2Fdynamicsurf.pdf&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10194974%2F&rft.rights=open