Matches in SemOpenAlex for { <https://semopenalex.org/work/W3015000521> ?p ?o ?g. }
- W3015000521 abstract "The instability in GAN training has been a long-standing problem despite remarkable research efforts. We identify that instability issues stem from difficulties of performing feature matching with mini-batch statistics, due to a fragile balance between the fixed target distribution and the progressively generated distribution. In this work, we propose Feature Quantization (FQ) for the discriminator, to embed both true and fake data samples into a shared discrete space. The quantized values of FQ are constructed as an evolving dictionary, which is consistent with feature statistics of the recent distribution history. Hence, FQ implicitly enables robust feature matching in a compact space. Our method can be easily plugged into existing GAN models, with little computational overhead in training. We apply FQ to 3 representative GAN models on 9 benchmarks: BigGAN for image generation, StyleGAN for face synthesis, and U-GAT-IT for unsupervised image-to-image translation. Extensive experimental results show that the proposed FQ-GAN can improve the FID scores of baseline methods by a large margin on a variety of tasks, achieving new state-of-the-art performance." @default.
- W3015000521 created "2020-04-10" @default.
- W3015000521 creator A5003075563 @default.
- W3015000521 creator A5006435999 @default.
- W3015000521 creator A5007442135 @default.
- W3015000521 creator A5011979551 @default.
- W3015000521 creator A5047233371 @default.
- W3015000521 date "2020-04-05" @default.
- W3015000521 modified "2023-10-01" @default.
- W3015000521 title "Feature Quantization Improves GAN Training" @default.
- W3015000521 cites W2099471712 @default.
- W3015000521 cites W2108598243 @default.
- W3015000521 cites W2117539524 @default.
- W3015000521 cites W2124067153 @default.
- W3015000521 cites W2156097299 @default.
- W3015000521 cites W2183341477 @default.
- W3015000521 cites W2187089797 @default.
- W3015000521 cites W2295107390 @default.
- W3015000521 cites W2328111639 @default.
- W3015000521 cites W2593414223 @default.
- W3015000521 cites W2593768305 @default.
- W3015000521 cites W2775288145 @default.
- W3015000521 cites W2894919355 @default.
- W3015000521 cites W2902666438 @default.
- W3015000521 cites W2913399920 @default.
- W3015000521 cites W2949930907 @default.
- W3015000521 cites W2953030256 @default.
- W3015000521 cites W2962760235 @default.
- W3015000521 cites W2962770929 @default.
- W3015000521 cites W2962793481 @default.
- W3015000521 cites W2962879692 @default.
- W3015000521 cites W2963133080 @default.
- W3015000521 cites W2963373786 @default.
- W3015000521 cites W2963592272 @default.
- W3015000521 cites W2963623951 @default.
- W3015000521 cites W2963799213 @default.
- W3015000521 cites W2963800509 @default.
- W3015000521 cites W2963836885 @default.
- W3015000521 cites W2963890275 @default.
- W3015000521 cites W2963981733 @default.
- W3015000521 cites W2964024144 @default.
- W3015000521 cites W2964209830 @default.
- W3015000521 cites W2970475648 @default.
- W3015000521 cites W2991040477 @default.
- W3015000521 cites W2995910740 @default.
- W3015000521 cites W2996690341 @default.
- W3015000521 cites W3007166982 @default.
- W3015000521 cites W3035574324 @default.
- W3015000521 cites W3118608800 @default.
- W3015000521 cites W3159890710 @default.
- W3015000521 hasPublicationYear "2020" @default.
- W3015000521 type Work @default.
- W3015000521 sameAs 3015000521 @default.
- W3015000521 citedByCount "2" @default.
- W3015000521 countsByYear W30150005212020 @default.
- W3015000521 crossrefType "posted-content" @default.
- W3015000521 hasAuthorship W3015000521A5003075563 @default.
- W3015000521 hasAuthorship W3015000521A5006435999 @default.
- W3015000521 hasAuthorship W3015000521A5007442135 @default.
- W3015000521 hasAuthorship W3015000521A5011979551 @default.
- W3015000521 hasAuthorship W3015000521A5047233371 @default.
- W3015000521 hasConcept C11413529 @default.
- W3015000521 hasConcept C119857082 @default.
- W3015000521 hasConcept C127759330 @default.
- W3015000521 hasConcept C138885662 @default.
- W3015000521 hasConcept C153180895 @default.
- W3015000521 hasConcept C154945302 @default.
- W3015000521 hasConcept C2776401178 @default.
- W3015000521 hasConcept C2779803651 @default.
- W3015000521 hasConcept C28855332 @default.
- W3015000521 hasConcept C41008148 @default.
- W3015000521 hasConcept C41895202 @default.
- W3015000521 hasConcept C52622490 @default.
- W3015000521 hasConcept C76155785 @default.
- W3015000521 hasConcept C774472 @default.
- W3015000521 hasConcept C83665646 @default.
- W3015000521 hasConcept C94915269 @default.
- W3015000521 hasConceptScore W3015000521C11413529 @default.
- W3015000521 hasConceptScore W3015000521C119857082 @default.
- W3015000521 hasConceptScore W3015000521C127759330 @default.
- W3015000521 hasConceptScore W3015000521C138885662 @default.
- W3015000521 hasConceptScore W3015000521C153180895 @default.
- W3015000521 hasConceptScore W3015000521C154945302 @default.
- W3015000521 hasConceptScore W3015000521C2776401178 @default.
- W3015000521 hasConceptScore W3015000521C2779803651 @default.
- W3015000521 hasConceptScore W3015000521C28855332 @default.
- W3015000521 hasConceptScore W3015000521C41008148 @default.
- W3015000521 hasConceptScore W3015000521C41895202 @default.
- W3015000521 hasConceptScore W3015000521C52622490 @default.
- W3015000521 hasConceptScore W3015000521C76155785 @default.
- W3015000521 hasConceptScore W3015000521C774472 @default.
- W3015000521 hasConceptScore W3015000521C83665646 @default.
- W3015000521 hasConceptScore W3015000521C94915269 @default.
- W3015000521 hasLocation W30150005211 @default.
- W3015000521 hasOpenAccess W3015000521 @default.
- W3015000521 hasPrimaryLocation W30150005211 @default.
- W3015000521 hasRelatedWork W1493373688 @default.
- W3015000521 hasRelatedWork W2577749540 @default.
- W3015000521 hasRelatedWork W2896786261 @default.
- W3015000521 hasRelatedWork W2897003273 @default.