Matches in SemOpenAlex for { <https://semopenalex.org/work/W4313051036> ?p ?o ?g. }
- W4313051036 abstract "Transformers have achieved great success in pluralistic image inpainting recently. However, we find existing transformer based solutions regard each pixel as a token, thus suffer from information loss issue from two aspects: 1) They downsample the input image into much lower resolutions for efficiency consideration, incurring information loss and extra misalignment for the boundaries of masked regions. 2) They quantize 256 <sup xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>3</sup> RGB pixels to a small number (such as 512) of quantized pixels. The indices of quantized pixels are used as tokens for the inputs and prediction targets of transformer. Although an extra CNN network is used to upsample and refine the low-resolution results, it is difficult to retrieve the lost information back. To keep input information as much as possible, we propose a new transformer based framework “PUT”. Specifically, to avoid input downsampling while maintaining the computation efficiency, we design a patch-based auto-encoder P-VQVAE, where the encoder converts the masked image into non-overlapped patch tokens and the decoder recovers the masked regions from the inpainted tokens while keeping the unmasked regions unchanged. To eliminate the information loss caused by quantization, an Un-Quantized Transformer (UQ-Transformer) is applied, which directly takes the features from P-VQVAE encoder as input without quantization and regards the quantized tokens only as prediction targets. Extensive experiments show that PUT greatly outperforms state-of-the-art methods on image fidelity, especially for large masked regions and complex large-scale datasets." @default.
- W4313051036 created "2023-01-06" @default.
- W4313051036 creator A5029523095 @default.
- W4313051036 creator A5033332385 @default.
- W4313051036 creator A5036850912 @default.
- W4313051036 creator A5045938154 @default.
- W4313051036 creator A5046214153 @default.
- W4313051036 creator A5057293861 @default.
- W4313051036 creator A5058724252 @default.
- W4313051036 creator A5064573190 @default.
- W4313051036 creator A5077341571 @default.
- W4313051036 date "2022-06-01" @default.
- W4313051036 modified "2023-10-13" @default.
- W4313051036 title "Reduce Information Loss in Transformers for Pluralistic Image Inpainting" @default.
- W4313051036 cites W1993120651 @default.
- W4313051036 cites W1999360130 @default.
- W4313051036 cites W2093212899 @default.
- W4313051036 cites W2105038642 @default.
- W4313051036 cites W2108598243 @default.
- W4313051036 cites W2475287302 @default.
- W4313051036 cites W2732026016 @default.
- W4313051036 cites W2738588019 @default.
- W4313051036 cites W2962770929 @default.
- W4313051036 cites W2963255313 @default.
- W4313051036 cites W2963420272 @default.
- W4313051036 cites W2982763192 @default.
- W4313051036 cites W2991377405 @default.
- W4313051036 cites W3034482833 @default.
- W4313051036 cites W3034577585 @default.
- W4313051036 cites W3035002246 @default.
- W4313051036 cites W3035637413 @default.
- W4313051036 cites W3048765086 @default.
- W4313051036 cites W3103174683 @default.
- W4313051036 cites W3107942585 @default.
- W4313051036 cites W3158252298 @default.
- W4313051036 cites W3167536469 @default.
- W4313051036 cites W3180355996 @default.
- W4313051036 cites W3190492058 @default.
- W4313051036 cites W3199003182 @default.
- W4313051036 cites W3214586131 @default.
- W4313051036 cites W3216270236 @default.
- W4313051036 cites W4221111246 @default.
- W4313051036 cites W4240726888 @default.
- W4313051036 cites W4249724384 @default.
- W4313051036 doi "https://doi.org/10.1109/cvpr52688.2022.01106" @default.
- W4313051036 hasPublicationYear "2022" @default.
- W4313051036 type Work @default.
- W4313051036 citedByCount "18" @default.
- W4313051036 countsByYear W43130510362022 @default.
- W4313051036 countsByYear W43130510362023 @default.
- W4313051036 crossrefType "proceedings-article" @default.
- W4313051036 hasAuthorship W4313051036A5029523095 @default.
- W4313051036 hasAuthorship W4313051036A5033332385 @default.
- W4313051036 hasAuthorship W4313051036A5036850912 @default.
- W4313051036 hasAuthorship W4313051036A5045938154 @default.
- W4313051036 hasAuthorship W4313051036A5046214153 @default.
- W4313051036 hasAuthorship W4313051036A5057293861 @default.
- W4313051036 hasAuthorship W4313051036A5058724252 @default.
- W4313051036 hasAuthorship W4313051036A5064573190 @default.
- W4313051036 hasAuthorship W4313051036A5077341571 @default.
- W4313051036 hasBestOaLocation W43130510362 @default.
- W4313051036 hasConcept C110384440 @default.
- W4313051036 hasConcept C111919701 @default.
- W4313051036 hasConcept C115961682 @default.
- W4313051036 hasConcept C11727466 @default.
- W4313051036 hasConcept C118505674 @default.
- W4313051036 hasConcept C119599485 @default.
- W4313051036 hasConcept C127413603 @default.
- W4313051036 hasConcept C154945302 @default.
- W4313051036 hasConcept C160633673 @default.
- W4313051036 hasConcept C165801399 @default.
- W4313051036 hasConcept C28855332 @default.
- W4313051036 hasConcept C31972630 @default.
- W4313051036 hasConcept C38652104 @default.
- W4313051036 hasConcept C41008148 @default.
- W4313051036 hasConcept C48145219 @default.
- W4313051036 hasConcept C66322947 @default.
- W4313051036 hasConceptScore W4313051036C110384440 @default.
- W4313051036 hasConceptScore W4313051036C111919701 @default.
- W4313051036 hasConceptScore W4313051036C115961682 @default.
- W4313051036 hasConceptScore W4313051036C11727466 @default.
- W4313051036 hasConceptScore W4313051036C118505674 @default.
- W4313051036 hasConceptScore W4313051036C119599485 @default.
- W4313051036 hasConceptScore W4313051036C127413603 @default.
- W4313051036 hasConceptScore W4313051036C154945302 @default.
- W4313051036 hasConceptScore W4313051036C160633673 @default.
- W4313051036 hasConceptScore W4313051036C165801399 @default.
- W4313051036 hasConceptScore W4313051036C28855332 @default.
- W4313051036 hasConceptScore W4313051036C31972630 @default.
- W4313051036 hasConceptScore W4313051036C38652104 @default.
- W4313051036 hasConceptScore W4313051036C41008148 @default.
- W4313051036 hasConceptScore W4313051036C48145219 @default.
- W4313051036 hasConceptScore W4313051036C66322947 @default.
- W4313051036 hasLocation W43130510361 @default.
- W4313051036 hasLocation W43130510362 @default.
- W4313051036 hasOpenAccess W4313051036 @default.
- W4313051036 hasPrimaryLocation W43130510361 @default.
- W4313051036 hasRelatedWork W1504109132 @default.
- W4313051036 hasRelatedWork W2117562399 @default.
- W4313051036 hasRelatedWork W2213520135 @default.