Automatic Correction of Internal Units in Generative Neural Networks

Tousi, Ali; Jeong, Haedong; Han, J.; Choi, Hwanil; Choi, Jaesik

doi:10.1109/cvpr46437.2021.00784

Cited by 4 publications

(5 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Artifact Correction To validate that the locally activated neurons are related to the artifact, we perform an ablation study on the High CLA group as described in (Tousi et al 2021). Instead of training an external classifier to identify the artifact causing internal units, we use the average CLA over the neurons in each unit.…”

Section: Qualitative Resultsmentioning

confidence: 99%

“…A sampling method with the trained generative boundaries was suggested to explain shared semantic information in the generator (Jeon, Jeong, and Choi 2020). Classifier-based defective internal featuremap unit identification was devised (Tousi et al 2021). The authors increase the visual fidelity by sequentially controlling the generation flow of the identified units.…”

Section: Related Workmentioning

confidence: 99%

“…In this section, we present our main contribution, the concept of local activation and its relation with low visual fidelity for individual generations. From previous research (Bau et al 2019;Jeon, Jeong, and Choi 2020;Tousi et al 2021), we can presume that each internal featuremap unit in the generator handles a specific object (e.g., tree, glasses) for the final generation. In particular, an artifact that has low visual fidelity can also be considered as a type of object.…”

Section: Locally Activated Neurons In Gansmentioning

confidence: 99%

“…The authors further improve the fidelity of individual generations by zero-ablating the detected featuremap units. A similar approach trains an external classifier to extract the region with low visual fidelity in the individual generations and identifies internal units related to the extracted region (Tousi et al 2021). On the other hand, manipulation of the latent code based on the binary linear classifier has been proposed to correct the artifact (Shen et al 2020).…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

An Unsupervised Way to Understand Artifact Generating Internal Units in Generative Neural Networks

Jeong

Han

Choi

2022

AAAI

Self Cite

View full text Add to dashboard Cite

Despite significant improvements on the image generation performance of Generative Adversarial Networks (GANs), generations with low visual fidelity still have been observed. As widely used metrics for GANs focus more on the overall performance of the model, evaluation on the quality of individual generations or detection of defective generations is challenging. While recent studies try to detect featuremap units that cause artifacts and evaluate individual samples, these approaches require additional resources such as external networks or a number of training data to approximate the real data manifold. In this work, we propose the concept of local activation, and devise a metric on the local activation to detect artifact generations without additional supervision. We empirically verify that our approach can detect and correct artifact generations from GANs with various datasets. Finally, we discuss a geometrical analysis to partially reveal the relation between the proposed concept and low visual fidelity.

show abstract

Section: Qualitative Resultsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Locally Activated Neurons In Gansmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

An Unsupervised Way to Understand Artifact Generating Internal Units in Generative Neural Networks

Jeong

Han

Choi

2022

AAAI

Self Cite

View full text Add to dashboard Cite

show abstract

“…As far as we know, only a few works have been proposed to deal with the artifacts of generative models. Two works [14], [40] have been conducted to detect artifacts produced by GAN models. However, these models are model-specific since they are limited to the latent space of GAN.…”

Section: Introductionmentioning

confidence: 99%