?url_ver=Z39.88-2004&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Adc&rft.title=A+Multi-Implicit+Neural+Representation+for+Fonts&rft.creator=Reddy%2C+P&rft.creator=Zhang%2C+Z&rft.creator=Wang%2C+Z&rft.creator=Fisher%2C+M&rft.creator=Jin%2C+H&rft.creator=Mitra%2C+NJ&rft.description=Fonts+are+ubiquitous+across+documents+and+come+in+a+variety+of+styles.+They+are+either+represented+in+a+native+vector+format+or+rasterized+to+produce+fixed+resolution+images.+In+the+first+case%2C+the+non-standard+representation+prevents+benefiting+from+latest+network+architectures+for+neural+representations%3B+while%2C+in+the+latter+case%2C+the+rasterized+representation%2C+when+encoded+via+networks%2C+results+in+loss+of+data+fidelity%2C+as+font-specific+discontinuities+like+edges+and+corners+are+difficult+to+represent+using+neural+networks.+Based+on+the+observation+that+complex+fonts+can+be+represented+by+a+superposition+of+a+set+of+simpler+occupancy+functions%2C+we+introduce+multi-implicits+to+represent+fonts+as+a+permutation-invariant+set+of+learned+implicit+functions%2C+without+losing+features+(e.g.%2C+edges+and+corners).+However%2C+while+multi-implicits+locally+preserve+font+features%2C+obtaining+supervision+in+the+form+of+ground+truth+multi-channel+signals+is+a+problem+in+itself.+Instead%2C+we+propose+how+to+train+such+a+representation+with+only+local+supervision%2C+while+the+proposed+neural+architecture+directly+finds+globally+consistent+multi-implicits+for+font+families.+We+extensively+evaluate+the+proposed+representation+for+various+tasks+including+reconstruction%2C+interpolation%2C+and+synthesis+to+demonstrate+clear+advantages+with+existing+alternatives.+Additionally%2C+the+representation+naturally+enables+glyph+completion%2C+wherein+a+single+characteristic+font+is+used+to+synthesize+a+whole+font+family+in+the+target+style.&rft.date=2021&rft.type=Proceedings+paper&rft.language=eng&rft.source=+++++In%3A++Advances+in+Neural+Information+Processing+Systems+34+(NeurIPS+2021).++(pp.+pp.+12637-12647).+++(2021)+++++&rft.format=text&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10151308%2F1%2FNeurIPS-2021-a-multi-implicit-neural-representation-for-fonts-Paper.pdf&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10151308%2F&rft.rights=open