EvolGAN: Evolutionary Generative Adversarial Networks

EvolGAN: Evolutionary Generative Adversarial Networks


We propose to use a quality estimator and evolutionary methods to search the latent space of generative adversarial networks trained on small, difficult datasets, or both. The new method leads to the generation of significantly higher quality images while preserving the original generator’s diversity. Human raters preferred an image from the new version with frequency 83.7% for Cats, 74% for FashionGen, 70.4% for Horses, and 69.2% for Artworks - minor improvements for the already excellent GANs for faces. This approach applies to any quality scorer and GAN generator.


9 \renewlistitemizeitemize5

1 Introduction

Generative adversarial networks (GAN) are the state-of-the-art generative models in many domains. However, they need quite a lot of training data to reach a decent performance. Using off-the-shelf image quality estimators, we propose a novel but simple evolutionary modification for making them more reliable for small, difficult, or multimodal datasets. Contrarily to previous approaches using evolutionary methods for image generation, we do not modify the training phase. We use a generator mapping a latent vector to an image built as in a classical GAN. The difference lies in the method used for choosing a latent vector . Instead of randomly generating a latent vector , we perform an evolutionary optimization, with as decision variables and the estimated quality of   based on a state-of-the-art quality estimation method as an objective function. We show that:

  • The quality of generated images is better, both for the proxy used for estimating the quality, i.e., the objective function, as well as for human raters. For example, the modified images are preferred by human raters more than 80% of the time for images of cats and around 70% of the time for horses and artworks.

  • The diversity of the original GAN is preserved: the new images are preferred by humans and still similar.

  • The computational overhead introduced by the evolutionary optimization is moderate, compared to the computational requirement for training the original GAN.

The approach is simple, generic, easy to implement , and fast. It can be used as a drop-in replacement for classical GAN provided that we have a quality estimator for the outputs of the GAN. Besides the training of the original GAN, many experiments were performed on a laptop without any GPU.

StyleGAN2 EvolGAN
Figure 1: For illustration , random images generated using StyleGAN2 (left) and EvolGAN (right). Horses were typically harder than cats. The images generated by EvolGAN are generally more realistic. The top-left example of a generated cat by StyleGAN2 has blood-like artifacts on its throat and the other is blurry. Three of the four StyleGAN2 horses are clearly unrealistic: on the bottom right of the StyleGan2 results, the human and the horse are mixed, the bottom left shows an incoherent mix of several horses, the top left looks like the ghost, and only the top right is realistic. Overall, both cats, and 3 of the 4 horses generated by EvolGAN look realistic. We show more examples of horses, as they are more difficult to model.
Figure 2: General EvolGAN approach (left) and optimization loop (right). Our method improves upon a pre-trained generator model G, which maps a latent vector to an image G(). In classical models, the images are generated by sampling a latent vector randomly. We modify that randomly chosen into a better evolved to improve the quality of the image, typically estimated by Koncept512. For preserving diversity, we ensure that is close to the original . The original image is generated using PokeGan (Pokemon dataset) and the improved one is generated using EvolGAN (superposed on top of PokeGan): we see an elephant-style Pokemon, hardly visible in the top version.

Fig. 1 shows examples of generations of compared to generations by StyleGAN2. Fig. 2 presents our general approach, detailed in the Section 3.

2 Related Works

2.1 Generative Adversarial Networks

Generative Adversarial Networks [?] (GANs) are widely used in machine learning [?, ?, ?, ?, ?, ?, ?, ?] for generative modeling. Generative Adversarial Networks are made up of two neural networks: a Generator , mapping a latent vector to an image and a Discriminator mapping an image to a realism value . Given a dataset , GANs are trained using two training steps operating concurrently:

  1. Given a randomly generated , the generator tries to fool into classifying its output as a real image, e.g. by maximizing . For this part of the training, only the weights of G are modified.

  2. Given a minibatch containing both random fake images and real images randomly drawn in , the discriminator learns to distinguish and , e.g. by optimizing the cross-entropy.

The ability of GANs to synthesize faces [6] is particularly impressive and of wide interest. However, such results are possible only with huge datasets for each modality and/or after careful cropping, which restricts their applicability. Here we consider the problem of improving GANs trained on small or difficult datasets. Classical tools for making GANs compatible with small datasets include:

  • Data augmentation, by translation, rotation, symmetries, or other transformations.

  • Transfer from an existing GAN trained on another dataset to a new dataset [?].

  • Modification of the distribution in order to match a specific request as done in several papers. [9] modifies the training, using quality assessement as we do; however they modify the training whereas we modify inference. In the same vein, [17] works on scale disentanglement: it also works at training time. These works could actually be combined with ours. [8] generates images conditionally to a classifier output or conditionally to a captioning network output. [13] and [2] condition the generation to a playability criterion (estimated by an agent using the GAN output) or some high-level constraints. [1] uses a variational autoencoder (VAE), so that constraints can be added to the generation: they can add an attribute (e.g. black hair) and still take into account a realism criterion extracted from the VAE: this uses labels from the dataset. [12] uses disentanglement of the latent space for semantic face editing: the user can modify a specific latent variable. [18] allows image editing and manipulation: it uses projections onto the output domain.

  • Biasing the dataset. [7] augments the dataset by generating images with a distribution skewed towards the minority classes.

  • Learning a specific probability distribution, rather than using a predefined, for example Gaussian, distribution. Such a method is advocated in [?].

The latter is the closest to the present work in the sense that we stay close to the goal of the original GAN, i.e. modeling some outputs without trying to bias the construction towards some subset. However, whereas [?] learn a probability distribution on the fly while training the GAN, our approach learns a classical GAN and modifies, a posteriori, the probability distribution by considering a subdomain of the space of the latent variables in which images have better quality. We could work on an arbitrary generative model based on latent variables, not only GANs. As opposed to all previously mentioned works, we improve the generation, without modifying the target distribution and without using any side-information or handcrafted criterion - our ingredient is a quality estimator. Other combinations of deep learning and evolutionary algorithms have been published around GANs. For instance, [14] evolves a population of generators, whereas our evolutionary algorithm evolves individuals in the latent space. [?] also evolves individuals in the latent space, but using human feedback rather than the quality estimators that we are using. [11] evolves individuals in the latent space, but either guided by human feedback or by using similarity to a target image.

2.2 Quality estimators: Koncept512 and AVA

Quality estimation is a long-standing research topic [15, 16] recently improved by deep learning [4]. In the present work, we focus on such quality estimation tools based on supervised convolutional networks. The KonIQ-10k dataset is a large publicly available image quality assessment dataset with 10,073 images rated by humans. Each image is annotated by 120 human raters. The Koncept512 image quality scorer [4] is based on an InceptionResNet-v2 architecture and trained on KonIQ-10k for predicting the mean opinion scores of the annotators. It takes as input an image and outputs a quality estimate . Koncept512 is the state of the art in technical quality estimation [4], and is freely available. We use the release without any modification. [3] provides a tool similar to Koncept512, termed AVA, but dedicated to aesthetics rather than technical quality. It was easy to apply it as a drop-in replacement of Koncept512 in our experiments.

3 Methods

3.1 Our algorithm: EvolGAN

We do not modify the training of the GAN. We use a generator created by a GAN. takes as input a random latent vector , and outputs an image . While the latent vector is generally chosen randomly (e.g., ), we treat it as a free parameter to be optimized according to a quality criterion . More formally, we obtain :


In this paper, is either Koncept512 or AVA. Our algorithm computes an approximate solution of problem 1 and outputs . Importantly, we do not want a global optimum of Eq. 1. We want a local optimum, in order to have essentially the same image must be close to , which would not happen without this condition. The optimization algorithm used to obtain in Eq. 1 is a simple -Evolution Strategy with random mutation rates [?], adapted as detailed in Section 3.2 (see Alg. 1). We keep the budget of our algorithm low, and the mutation strength parameter can be used to ensure that the image generated by EvolGAN is similar to the initial image. For instance, with , the expected number of mutated variables is, by construction (see Section 3.1), bounded by . We sometimes use the aesthetic quality estimator AVA rather than the technical quality estimator Koncept512 for quality estimation. We consider a coordinate-wise mutation rate: we mutate or do not mutate each coordinate, independently with some probability.

3.2 Optimization algorithms

After a few preliminary trials we decided to use the -Evolution Strategy with uniform mixing of mutation rates [?], with a modification as described in algorithm 1. This modification is designed for tuning the compromise between quality and diversity as discussed in Table 1.

Parameters : 
  • A probability distribution on .

  • A quality estimator , providing an estimate of the quality of some . We use or .

  • A generator , building for .

  • A budget .

  • A mutation strength .

  • A randomly generated . is the baseline image we are trying to improve.

  • 1 for  do
    4       for  do
    5             with probability , ( marginal of ).
    6       end for
    7      if ) then
    9       end if
    11 end for
    Output :  Optimized image
    Algorithm 1 The algorithm

    We used . Optionally, can be provided as an argument, leading to . The difference with the standard uniform mixing of mutation rates is that . With , the resulting image is close to the original image , whereas with the outcome is not similar to . Choosing (or , closely related to FastGA[?]) leads to faster convergence rates but also to less diversity (see Alg.1, line 1). We will show that overall, is the best choice for EvolGAN. We therefore get algorithms as presented in Table 1.

    behavior of with budget
    standard evol. alg.
    with mutation rate .
    uniform mixing of mutation rates
    [?] (also related to [?]).
    all variables mutated: equivalent
    to random search
    intermediate values intermediate behavior
    Table 1: Optimization algorithms used in the present paper. The last setting is new compared to [?]. We modified the maximum mutation rate for doing a local or global search depending on , so that the diversity of the outputs is maintained when is small (Sect. 4.4). denotes the number of non-zero components of .

    3.3 Open source codes

    We use the GAN publicly available at https://github.com/moxiegushi/pokeGAN, which is an implementation of Wasserstein GAN [?], the StyleGAN2 [6] available at thispersondoesnotexist.com, and PGAN on FashionGen from Pytorch GAN zoo [10]. Koncept512 is available at https://github.com/subpic/koniq. Our combination of Koncept512 and PGAN is available at DOUBLEBLIND. We use the evolutionary programming platform Nevergrad [?].

    4 Experiments

    We present applications of EvolGAN on three different GAN models: (i) StyleGAN2 for faces, cats, horses and artworks (ii) PokeGAN for mountains and Pokemons (iii) PGAN from Pytorch GAN zoo for FashionGen.

    4.1 Quality improvement on StyleGAN2

    The experiments are based on open source codes [?, 10, ?, 4, 3]. We use the StyleGAN2 [6] trained on a horse dataset, a face dataset, an artwork dataset, and a cat dataset1. Faces. We conducted a human study to assess the quality of EvolGAN compared to StyleGAN, by asking to 10 subject their preferred generations (pairwise comparisons, double-blind, random positioning). There were 11 human raters in total, including both experts with a strong photography background and beginners. 70% of the ratings came from experts. Results appear in Table 2. Faces are the most famous result of StyleGAN2. Although the results are positive as the images generated by EvolGAN are preferred significantly more than 50% of the time, the difference between StyleGAN2 and EvolGAN is quite small on this essentially solved problem compared to wild photos of cats or horses or small datasets.

    50.0 57.1
    75.0 44.4 66.7
    53.8 53.8 40.0 46.2
    10-80 ag-
    gregated 60.4% 3.4% (208 ratings)
    Table 2: Human study on faces dataset. , quality estimator . Row , col. : frequency at which human raters preferred to . By construction, for all , is equal to the original GAN. The fifth row aggregates all results of the first four rows for more significance.
    Dataset Budget Quality estimator score
    Cats 300 Koncept512 83.71 1.75% (446 ratings)
    Horses 300 Koncept512 70.43 4.27% (115 ratings)
    Artworks 300 Koncept512 69.19 3.09% (224 ratings)
    Table 3: Difficult test beds. ; same protocol as in Tables 2 i.e. we check with which probability human raters prefer an image generated by to an image generated by the original Gan . By definition of EvolGAN, . Number are above : using EvolGAN for modifying the latent vector improves the original StyleGAN2.
    Context LPIPS score
    PGAN 0.306 0.0003
    0.303 0.0003
    0.286 0.0003
    0.283 0.0002
    Table 4: LPIPS scores on FashionGen. As expected, mostly preserves the diversity of the generated images, while higher values of can lead to less diversity for the output of EvolGAN. The LPIPS was computed on samples of images for each setting.

    Harder settings. Animals and artworks are a much more difficult setting (Fig. 3) - StyleGAN2 sometimes fails to propose a high quality image. Fig. 3 presents examples of generations of and in such cases. Here, has more headroom for providing improvements than for faces: results are presented in Table 3. The case of horses or cats is particularly interesting: the failed generations often contain globally unrealistic elements, such as random hair balls flying in the air or unusual positioning of limbs, which are removed by EvolGAN. For illustration purpose, in Fig. 3 we present a few examples of generations which go wrong for the original StyleGan2 and for ; the bad examples in the case of the original StyleGan2 are much worse.

    Figure 3: For illustration, bad generations by StyleGAN2 and by EvolGAN. (a) Generation of a cat by StyleGAN2: we looked for a bad generation and found that one. Such bad cases completely disappear in the EvolGAN counterpart. (b) example of cat generation by EvolGAN: we failed to find a really bad EvolGAN generated image. (c) Example of bad horse generation by EvolGAN: the shape is unusual (looks like the muzzle of a pork) but we still recognize a horse. (d) Bad generation of a horse by StyleGAN2: some hair balls are randomly flying in the air.

    4.2 Small difficult datasets and

    In this section we focus on the use of EvolGAN for small datasets. We use the original pokemon dataset in PokeGAN [?] and an additional dataset created from copyright-free images of mountains. The previous section was quite successful, using (i.e. random search). The drawback is that the obtained images are not necessarily related to the original ones, and we might lose diversity (though Section 4.4 shows that this is not always the case, see discussion later). We will see that fails in the present case. In this section, we use small, and check if the obtained images are better than (see Table 5) and close to the original image (see Fig.4). Fig. 4 presents a Pokemon generated by the default GAN and its improved counterpart obtained by with . Table 5 presents our experimental setting and the results of our human study conducted on PokeGAN. We see a significant improvement when using Koncept512 on real-world data (as opposed to drawings such as Pokemons, for which Koncept512 fails), whereas we fail with AVA as in previous experiments (see Table 2). We succeed on drawings with Koncept512 only with : on this dataset of drawings (poorly adapted to Koncept512), large leads to a pure black image.

    Type Number of Number of Budget Quality Frequency
    of of training estimator of image
    images images epochs preferred to original
    Real world scenes
    Mountains 84 4900 500 Koncept512 73.3% 4.5% (98 ratings)
    Artificial scenes
    Pokemons 1840 4900 500 Koncept512 55%
    Pokemons 1840 4900 2000 Koncept512 52%
    Pokemons 1840 4900 6000 Koncept512 56.3 5.2% (92 ratings)
    Artificial scenes, higher mutation rates
    Pokemons 1840 4900 500 Koncept512 36.8%
    Pokemons 1840 4900 20 Koncept512 0%
    Table 5: Experimental results with . Reading guide: the last column shows the probability that an image was prefered to the starting point . The dimension of the latent space is except for mountains (). Koncept512 performs well on real world scenes but not on artificial scenes. For Pokemon with , the 0% (0 success out of 24 tests!) is interesting: the code starts to generate almost uniform images even with a budget , showing that Koncept512 fails on drawings. On mountains (the same GAN, but trained on real world images instead of Pokemons), and to a lesser extent on Pokemons for small , the images generated using EvolGAN are preferred more than of the time: using EvolGAN for modifying the latent vector improves the original PokeGan network.
    Figure 4: In both cases, a Pokemon generated by the default GAN (left) and after improvement by Koncept512 (right). For the left pair, after improvement, we see eyes and ears for a small elephant-style pokemon sitting on his back. A similar transformation appears for the more rabbit-style pokemon on the right hand side. These cherry-picked examples (cherry-picked, i.e. we selected specific cases for illustration purpose) are, however, less convincing than the randomly generated examples in Fig. 5 - Pokemons are the least successful applications, as Koncept512, with large or big budgets, tends to push those artificial images towards dark almost uniform images.

    4.3 Quality improvement

    Pytorch Gan Zoo [10] is an implementation of progressive GANs (PGAN[?]), applied here with FashionGen [?] as a dataset. The dimension of the latent space is 256. In Table 6, we present the results of our human study comparing to . With , humans prefer EvolGAN to the baseline in more than 73% of cases, even after only 40 iterations. also ensures that the images stay close to the original images when the budget is low enough (see Table 1). Fig. 5 shows some examples of generations using EvolGAN and the original PGAN.

    b Frequency
    40 73.33 8.21% (30 ratings))
    320 75.00 8.33% (28 ratings))
    40 and 320 aggreg. 74.13 5.79% (58 ratings))
    40 48.27 9.44% (29 ratings)
    320 67.74 8.53% (31 ratings)
    40 and 320 aggreg. 58.33 6.41% (60 ratings)
    40 56.66 9.20% (30 ratings)
    320 66.66 9.24% (27 ratings)
    40 and 320 aggreg. 61.40 6.50% (57 ratings)
    All aggregated
    40 59.55 5.23% (89 ratings)
    320 69.76 4.98% (86 ratings)
    40 and 320 aggreg. 64.57 3.62% (175 ratings)
    Table 6: Frequency (a.k.a score) at which various versions of were preferred to their starting point on the FashionGen dataset. This experiment is performed with Koncept512 as a quality estimator. In most experiments we get the best results with and do not need more than a budget . The values are greater than , meaning that EvolGAN improves the original PGAN network on FashionGen according to human preferences.
    Figure 5: Preservation of diversity, in particular with budget , when . We present triplets , i.e. in each case the output of PGAN and its optimized counterpart with budgets 40 and 320 respectively (images 1, 4, 7, 10, 13, 16, 19, 22 are the and images 2, 5, 8, 11, 14, 17, 20, 23 are their counterparts for budget ; indices +1 for budget 320). With (unpresented) and are quite different, so that we can not guarantee that diversity is preserved. With diversity is preserved with : for each of these 8 cases, the second image () is quite close to the original image, just technically better — except the one for which is quite buggy and EvolGAN can rightly move farther from the original image. Diversity is less preserved with : e.g. on the top right we see that the dress becomes shorter at .

    4.4 Consistency: preservation of diversity.

    Here we show that the generated image is close to the one before the optimization. More precisely, given , the following two methods provide related outputs: method 1 (classical GAN) outputs , and method 2 (EvolGAN) outputs , where is obtained by our evolutionary algorithms starting at with budget and parameter (Sect. 3.2). Fig. 5 shows some example generated images using PGAN and EvolGAN. For most examples, is very similar to so the diversity of the original GAN is preserved well. Following [5, 19], we measure numerically the diversity of the generated images from the PGAN model, and from EvolGAN models based on it, using the LPIPS score. The scores were computed on samples of images generated with each method. For each sample, we computed the LPIPS with another randomly-chosen generated image. The results are presented in Table 4. Higher values correspond to higher diversity of samples. EvolGAN preserves the diversity of the images generated when used with .

    4.5 Using AVA rather than Koncept512

    In Table 7 we show that AVA is less suited than Koncept512  as a quality assessor in EvolGAN. The human annotators do not find the images generated using EvolGAN with AVA to be better than those generated without EvolGAN. We hypothesize that this is due to the subjectivity of what AVA estimates: aesthetic quality. While humans generally agree on the factors accounting for the technical quality of an image (sharp edges, absence of blur, right brightness), annotators often disagree on aesthetic quality. Another factor may be that aesthetics are inherently harder to evaluate than technical quality.

    52.0 42.8
    39.1 32.0 36.4
    52.2 52.2 40.9 56.0%
    44.5% 5.0%
    500 50.55 3.05 %

    (a) Faces with StyleGAN2: reproducing Table 2 with AVA in lieu of Koncept512.
    Dataset Budget Quality estimator score Cats 300 AVA 47.05 7.05% Artworks 300 AVA 55.71 5.97 %
    (b) Reproducing Table 3 with AVA in lieu of Koncept512.
    Type Number of Number of Budget Quality Frequency of of training estimator of image images images epochs preferred to original Mountains 84 4900 500 AVA 42.5% Pokemons 1840 4900 500 AVA 52.6% Pokemons 1840 4900 500 AVA 52.6%
    (c) Reproducing Table 5 with AVA rather than Koncept512.

    Table 7: Testing AVA rather than Koncept512 as a quality estimator. With AVA, EvolGAN fails to beat the baseline according to human annotators.

    5 Conclusion

    We have shown that, given a generative model , optimizing by an evolutionary algorithm using Koncept512 as a criterion and preferably with (i.e. the classical -Evolution Strategy), leads to

    • The generated images are preferred by humans as shown on Table 3 for StyleGAN2, Table 5 for PokeGan and Table 6 for PGAN on FashionGen

    • The diversity is preserved, as shown on Fig. 5 and Table 4, when using a small value for (see Section 3.2).

    Choosing . small, i.e., the classical -Evolution Strategy with mutation rate , is usually the best choice: we preserve the diversity (with provably a small number of mutated variables, and experimentally a resulting image differing from the original one mainly in terms of quality), and the improvement compared to the original GAN is clearer as we can directly compare to  — a budget was usually enough. Importantly, evolutionary algorithms clearly outperformed random search and not only in terms of speed: we completely lose the diversity with random search, as well as the ability to improve a given point. Our application of evolution is a case in which we provably preserve diversity — with a mutation rate bounded by , and a budget , and a dimension , we get an expected ratio of mutated variables at most . In our setting so the maximum expected ratio of mutated variables is at most in Fig. 5. A tighter, run-dependent bound, can be obtained by comparing and and could be a stopping criterion.

    Successes. We get an improved GAN without modifying the training. The results are positive in all cases in particular difficult real-world data (Table 3), though the gap is moderate when the original model is already excellent (faces, Table 2) or when data are not real-world (Pokemons, Table 5.EvolGAN with Koncept512 is particularly successful on several difficult cases with real-world data — Mountains with Pokegan, Cats, Horses and Artworks with StyleGAN2 and FashionGen with Pytorch Gan Zoo.

    Remark on quality assessement. Koncept512 can be used on a wide range of applications. As far as our framework can tell, it outperforms AVA as a tool for EvolGAN (Table 7). However, it fails on artificial scenes such as Pokemons, unless, we use a small for staying local.

    Computational cost. All the experiments with PokeGAN presented here could be run on a laptop without using any GPU. The experiments with StyleGAN2 and PGAN use at most 500 (and often just 40) calls to the original GAN, without any specific retraining: we just repeat the inference with various latent vectors chosen by our evolutionary algorithm as detailed in Section 3.1.


    Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation), Project-ID 251654672, TRR 161 (Project A05).


    1. https://www.thishorsedoesnotexist.com/,https://www.thispersondoesnotexist.com/,https://www.thisartworkdoesnotexist.com/,https://www.thiscatdoesnotexist.com/


    1. J. H. Engel, M. Hoffman and A. Roberts (2017) Latent constraints: learning to generate conditionally from unconditional generative models. CoRR abs/1711.05772. External Links: 1711.05772 Cited by: 3rd item.
    2. E. Giacomello, P. L. Lanzi and D. Loiacono (2019) Searching the latent space of a generative adversarial network to generate doom levels. In 2019 IEEE Conference on Games (CoG), Vol. , pp. 1–8. Cited by: 3rd item.
    3. V. Hosu, B. Goldlucke and D. Saupe (2019) Effective aesthetics prediction with multi-level spatially pooled features. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 9375–9383. Cited by: §2.2, §4.1.
    4. V. Hosu, H. Lin, T. Sziranyi and D. Saupe (2020) KonIQ-10k: an ecologically valid database for deep learning of blind image quality assessment. IEEE Transactions on Image Processing (), pp. 1–1. External Links: Document, ISSN 1941-0042 Cited by: §2.2, §4.1.
    5. X. Huang, M. Liu, S. J. Belongie and J. Kautz (2018) Multimodal unsupervised image-to-image translation. CoRR abs/1804.04732. External Links: 1804.04732 Cited by: §4.4.
    6. T. Karras, S. Laine, M. Aittala, J. Hellsten, J. Lehtinen and T. Aila (2019) Analyzing and improving the image quality of stylegan. External Links: 1912.04958 Cited by: §2.1, §3.3, §4.1.
    7. G. Mariani, F. Scheidegger, R. Istrate, C. Bekas and C. Malossi (2018) Bagan: data augmentation with balancing gan. arXiv preprint arXiv:1803.09655. Cited by: 4th item.
    8. A. Nguyen, J. Clune, Y. Bengio, A. Dosovitskiy and J. Yosinski (2016) Plug & play generative networks: conditional iterative generation of images in latent space. External Links: 1612.00005 Cited by: 3rd item.
    9. K. PARIMALA and S. Channappayya (2019) Quality aware generative adversarial networks. In Advances in Neural Information Processing Systems, pp. 2948–2958. Cited by: 3rd item.
    10. M. Riviere (2019) Pytorch GAN Zoo. GitHub. Note: \urlhttps://GitHub.com/FacebookResearch/pytorch_GAN_zoo Cited by: §3.3, §4.1, §4.3.
    11. M. Riviere, O. Teytaud, J. Rapin, Y. LeCun and C. Couprie (2019) Inspirational adversarial image generation. arXiv preprint 1906.11661. Cited by: §2.1.
    12. Y. Shen, J. Gu, X. Tang and B. Zhou (2019) Interpreting the latent space of gans for semantic face editing. External Links: 1907.10786 Cited by: 3rd item.
    13. V. Volz, J. Schrum, J. Liu, S. M. Lucas, A. Smith and S. Risi (2018) Evolving mario levels in the latent space of a deep convolutional generative adversarial network. In Proceedings of the Genetic and Evolutionary Computation Conference, GECCO ’18, New York, NY, USA, pp. 221–228. External Links: ISBN 9781450356183, Document Cited by: 3rd item.
    14. C. Wang, C. Xu, X. Yao and D. Tao (2018) Evolutionary generative adversarial networks. CoRR abs/1803.00657. External Links: 1803.00657 Cited by: §2.1.
    15. Z. Wang, A. C. Bovik, H. R. Sheikh and E. P. Simoncelli (2004) Image quality assessment: from error visibility to structural similarity. IEEE Transactions on Image Processing. Cited by: §2.2.
    16. P. Ye, J. Kumar, L. Kang and D. Doermann (2012) Unsupervised feature learning framework for no-reference image quality assessment. IEEE Conference on Computer Vision and Pattern Recognition. Cited by: §2.2.
    17. Z. Yi, Z. Chen, H. Cai, W. Mao, M. Gong and H. Zhang (2020) BSD-gan: branched generative adversarial network for scale-disentangled representation learning and image synthesis. IEEE Transactions on Image Processing. Cited by: 3rd item.
    18. J. Zhu, P. Krähenbühl, E. Shechtman and A. A. Efros (2016) Generative visual manipulation on the natural image manifold. In European conference on computer vision, pp. 597–613. Cited by: 3rd item.
    19. J. Zhu, R. Zhang, D. Pathak, T. Darrell, A. A. Efros, O. Wang and E. Shechtman (2017) Toward multimodal image-to-image translation. In Advances in neural information processing systems, pp. 465–476. Cited by: §4.4.
    Comments 0
    Request Comment
    You are adding the first comment!
    How to quickly get a good reply:
    • Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
    • Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
    • Your comment should inspire ideas to flow and help the author improves the paper.

    The better we are at sharing our knowledge with each other, the faster we move forward.
    The feedback must be of minimum 40 characters and the title a minimum of 5 characters
    Add comment
    Loading ...
    This is a comment super asjknd jkasnjk adsnkj
    The feedback must be of minumum 40 characters
    The feedback must be of minumum 40 characters

    You are asking your first question!
    How to quickly get a good answer:
    • Keep your question short and to the point
    • Check for grammar or spelling errors.
    • Phrase it like a question
    Test description