Matches in SemOpenAlex for { <https://semopenalex.org/work/W4214701208> ?p ?o ?g. }
- W4214701208 abstract "Generative Adversarial Networks (GANs) produce impressive results on unconditional image generation when powered with large-scale image datasets. Yet generated images are still easy to spot especially on datasets with high variance (e.g. bedroom, church). In this paper, we propose various improvements to further push the boundaries in image generation. Specifically, we propose a novel dual contrastive loss and show that, with this loss, discriminator learns more generalized and distinguishable representations to incentivize generation. In addition, we revisit attention and extensively experiment with different attention blocks in the generator. We find attention to be still an important module for successful image generation even though it was not used in the recent state-of-the-art models. Lastly, we study different attention architectures in the discriminator, and propose a reference attention mechanism. By combining the strengths of these remedies, we improve the compelling state-of-the-art Fréchet Inception Distance (FID) by at least 17.5% on several benchmark datasets. We obtain even more significant improvements on compositional synthetic scenes (up to 47.5% in FID)." @default.
- W4214701208 created "2022-03-02" @default.
- W4214701208 creator A5003887059 @default.
- W4214701208 creator A5025713692 @default.
- W4214701208 creator A5029174306 @default.
- W4214701208 creator A5033212845 @default.
- W4214701208 creator A5041525280 @default.
- W4214701208 creator A5055264952 @default.
- W4214701208 creator A5066242985 @default.
- W4214701208 date "2021-10-01" @default.
- W4214701208 modified "2023-09-25" @default.
- W4214701208 title "Dual Contrastive Loss and Attention for GANs" @default.
- W4214701208 cites W1834627138 @default.
- W4214701208 cites W1955055330 @default.
- W4214701208 cites W2108598243 @default.
- W4214701208 cites W2138621090 @default.
- W4214701208 cites W2145287260 @default.
- W4214701208 cites W2157364932 @default.
- W4214701208 cites W2171590421 @default.
- W4214701208 cites W2550553598 @default.
- W4214701208 cites W2561715562 @default.
- W4214701208 cites W2593414223 @default.
- W4214701208 cites W2601564443 @default.
- W4214701208 cites W2738588019 @default.
- W4214701208 cites W2752782242 @default.
- W4214701208 cites W2798991696 @default.
- W4214701208 cites W2953327099 @default.
- W4214701208 cites W2955058313 @default.
- W4214701208 cites W2962770929 @default.
- W4214701208 cites W2962974533 @default.
- W4214701208 cites W2963073614 @default.
- W4214701208 cites W2963091558 @default.
- W4214701208 cites W2963121829 @default.
- W4214701208 cites W2963312584 @default.
- W4214701208 cites W2963495494 @default.
- W4214701208 cites W2963800363 @default.
- W4214701208 cites W2963954913 @default.
- W4214701208 cites W2964024144 @default.
- W4214701208 cites W2964110616 @default.
- W4214701208 cites W2966926453 @default.
- W4214701208 cites W2979672901 @default.
- W4214701208 cites W2981721547 @default.
- W4214701208 cites W2982763192 @default.
- W4214701208 cites W2982795046 @default.
- W4214701208 cites W2983446232 @default.
- W4214701208 cites W2989855043 @default.
- W4214701208 cites W2991318208 @default.
- W4214701208 cites W3034515601 @default.
- W4214701208 cites W3034577585 @default.
- W4214701208 cites W3034600949 @default.
- W4214701208 cites W3034864980 @default.
- W4214701208 cites W3034885317 @default.
- W4214701208 cites W3034907931 @default.
- W4214701208 cites W3034954643 @default.
- W4214701208 cites W3035524453 @default.
- W4214701208 cites W3035574324 @default.
- W4214701208 cites W3035687950 @default.
- W4214701208 cites W3043547428 @default.
- W4214701208 cites W3094954720 @default.
- W4214701208 cites W3107096356 @default.
- W4214701208 cites W3119106047 @default.
- W4214701208 cites W3162384219 @default.
- W4214701208 cites W3187526215 @default.
- W4214701208 cites W3197074992 @default.
- W4214701208 doi "https://doi.org/10.1109/iccv48922.2021.00666" @default.
- W4214701208 hasPublicationYear "2021" @default.
- W4214701208 type Work @default.
- W4214701208 citedByCount "15" @default.
- W4214701208 countsByYear W42147012082021 @default.
- W4214701208 countsByYear W42147012082022 @default.
- W4214701208 countsByYear W42147012082023 @default.
- W4214701208 crossrefType "proceedings-article" @default.
- W4214701208 hasAuthorship W4214701208A5003887059 @default.
- W4214701208 hasAuthorship W4214701208A5025713692 @default.
- W4214701208 hasAuthorship W4214701208A5029174306 @default.
- W4214701208 hasAuthorship W4214701208A5033212845 @default.
- W4214701208 hasAuthorship W4214701208A5041525280 @default.
- W4214701208 hasAuthorship W4214701208A5055264952 @default.
- W4214701208 hasAuthorship W4214701208A5066242985 @default.
- W4214701208 hasBestOaLocation W42147012082 @default.
- W4214701208 hasConcept C115961682 @default.
- W4214701208 hasConcept C119857082 @default.
- W4214701208 hasConcept C121332964 @default.
- W4214701208 hasConcept C124952713 @default.
- W4214701208 hasConcept C13280743 @default.
- W4214701208 hasConcept C142362112 @default.
- W4214701208 hasConcept C153180895 @default.
- W4214701208 hasConcept C154945302 @default.
- W4214701208 hasConcept C163258240 @default.
- W4214701208 hasConcept C185798385 @default.
- W4214701208 hasConcept C205649164 @default.
- W4214701208 hasConcept C2779803651 @default.
- W4214701208 hasConcept C2780980858 @default.
- W4214701208 hasConcept C2780992000 @default.
- W4214701208 hasConcept C39890363 @default.
- W4214701208 hasConcept C41008148 @default.
- W4214701208 hasConcept C62520636 @default.
- W4214701208 hasConcept C76155785 @default.
- W4214701208 hasConcept C94915269 @default.
- W4214701208 hasConceptScore W4214701208C115961682 @default.