• 제목/요약/키워드: Defense-VAE

검색결과 3건 처리시간 0.014초

이미지 분할 여부에 따른 VQ-VAE 모델의 적대적 예제 복원 성능 비교 (Comparison of Adversarial Example Restoration Performance of VQ-VAE Model with or without Image Segmentation)

  • 김태욱;현승민;홍정희
    • 융합신호처리학회논문지
    • /
    • 제23권4호
    • /
    • pp.194-199
    • /
    • 2022
  • 다양하고 복잡한 영상 데이터 기반의 산업에서 높은 정확도와 활용성을 위해 고품질의 데이터를 위한 전처리가 요구된다. 하지만 기존 이미지 또는 영상 데이터와 노이즈를 결합해 기업에 큰 위험을 초래할 수 있는 오염된 적대적 예제가 유입될 시 기업의 신뢰도 및 보안성, 완전한 결과물 확보를 위해 손상되기 이전으로의 복원이 필요하다. 이를 위한 대비책으로 기존에는 Defense-GAN을 사용하여 복원을 진행하였지만, 긴 학습 시간과 복원물의 낮은 품질 등의 단점이 존재하였다. 이를 개선하기 위해 본 논문에서는 VQ-VAE 모델을 사용함과 더불어 이미지 분할 여부에 따라 FGSM을 통해 만든 적대적 예제를 이용하는 방법을 제안한다. 먼저, 생성된 예제를 일반 분류기로 분류한다. 다음으로 분할 전의 데이터를 사전 학습된 VQ-VAE 모델에 전달하여 복원한 후 분류기로 분류한다. 마지막으로 4등분으로 분할된 데이터를 4-split-VQ-VAE 모델에 전달하여 복원한 조각을 합친 뒤 분류기에 넣는다. 최종적으로 복원된 결과와 정확도를 비교한 후 분할 여부에 따른 2가지 모델의 결합 순서에 따라 성능을 분석한다.

Combination of red ginseng and velvet antler extracts prevents skin damage by enhancing the antioxidant defense system and inhibiting MAPK/AP-1/NF-κB and caspase signaling pathways in UVB-irradiated HaCaT keratinocytes and SKH-1 hairless mice

  • Van-Long Truong;Yeon-Ji Bae;Ji-Hong Bang;Woo-Sik Jeong
    • Journal of Ginseng Research
    • /
    • 제48권3호
    • /
    • pp.323-332
    • /
    • 2024
  • Background: Studies have reported that the combination of two or more therapeutic compounds at certain ratios has more noticeable pharmaceutical properties than single compounds and requires reduced dosage of each agent. Red ginseng and velvet antler have been extensively used in boosting immunity and physical strength and preventing diseases. Thus, this study was conducted to elucidate the skin-protective potentials of red ginseng extract (RGE) and velvet antler extract (VAE) alone or in combination on ultraviolet (UVB)-irradiated human keratinocytes and SKH-1 hairless mice. Methods: HaCaT cells were preincubated with RGE/VAE alone or in combination for 2 h before UVB (30 mJ/cm2) irradiation. SKH-1 mice were orally given RGE/VAE alone or in combination for 15 days before exposure to single dose of UVB (600 mJ/cm2). Treated cells and treated skin tissues were collected and subjected to subsequent experiments. Results: RGE/VAE pretreatment alone or in combination significantly prevented UVB-induced cell death, apoptosis, reactive oxygen species production, and DNA damage in keratinocytes and SKH-1 mouse skins by downregulating mitogen-activated protein kinases/activator protein 1/nuclear factor kappa B and caspase signaling pathways. These extracts also strengthened the antioxidant defense systems and skin barriers in UVB-irradiated HaCaT cells and SKH-1 mouse skins. Furthermore, RGE/VAE co-administration appeared to be more effective in preventing UVB-caused skin injury than these extracts used alone. Conclusion: Overall, these findings suggest that the consumption of RGE/VAE, especially in combination, offers a protective ability against UVB-caused skin injury by preventing inflammation and apoptosis and enhancing antioxidant capacity.

변이형 오토인코더를 이용한 탄도미사일 궤적 증강기법 개발 (Development of Augmentation Method of Ballistic Missile Trajectory using Variational Autoencoder)

  • 이동규;홍동욱
    • 시스템엔지니어링학술지
    • /
    • 제19권2호
    • /
    • pp.145-156
    • /
    • 2023
  • Trajectory of ballistic missile is defined by inherent flight dynamics, which decided range and maneuvering characteristics. It is crucial to predict range and maneuvering characteristics of ballistic missile in KAMD (Korea Air and Missile Defense) to minimize damage due to ballistic missile attacks, Nowadays, needs for applying AI(Artificial Intelligence) technologies are increasing due to rapid developments of DNN(Deep Neural Networks) technologies. To apply these DNN technologies amount of data are required for superviesed learning, but trajectory data of ballistic missiles is limited because of security issues. Trajectory data could be considered as multivariate time series including many variables. And augmentation in time series data is a developing area of research. In this paper, we tried to augment trajectory data of ballistic missiles using recently developed methods. We used TimeVAE(Time Variational AutoEncoder) method and TimeGAN(Time Generative Adversarial Networks) to synthesize missile trajectory data. We also compare the results of two methods and analyse for future works.