• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권6호
• /
• pp.2115-2127
• /
• 2021

Although significant progress has been made in synthesizing visually realistic face images by Generative Adversarial Networks (GANs), there still lacks effective approaches to provide fine-grained control over the generation process for semantic facial attribute editing. In this work, we propose a novel cross channel self-attention based generative adversarial network (CCA-GAN), which weights the importance of multiple channels of features and archives pixel-level feature alignment and conversion, to reduce the impact on irrelevant attributes while editing the target attributes. Evaluation results show that CCA-GAN outperforms state-of-the-art models on the CelebA dataset, reducing Fréchet Inception Distance (FID) and Kernel Inception Distance (KID) by 15~28% and 25~100%, respectively. Furthermore, visualization of generated samples confirms the effect of disentanglement of the proposed model.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제13권2호
• /
• pp.871-891
• /
• 2019

Visual tracking is a challenging task that needs learning an effective model to handle the changes of target appearance caused by factors such as pose variation, illumination change, occlusion and motion blur. In this paper, a novel tracking algorithm based on weighted local sub-space reconstruction error is presented. First, accounting for the appearance changes in the tracking process, a generative weight calculation method based on structural reconstruction error is proposed. Furthermore, a template update scheme of occlusion-aware is introduced, in which we reconstruct a new template instead of simply exploiting the best observation for template update. The effectiveness and feasibility of the proposed algorithm are verified by comparing it with some state-of-the-art algorithms quantitatively and qualitatively.

• 한국컴퓨터정보학회논문지
• /
• 제29권7호
• /
• pp.11-20
• /
• 2024

본 논문에서는 생성 모델을 이용한 데이터 프리 양자화에서 발생할 수 있는 지식 격차를 줄이기 위하여 BAG (Bit-width Aware Generator)와 채널 어텐션 기반 중간 레이어 지식 증류를 제안한다. 생성 모델을 이용한 데이터 프리 양자화의 생성자는 오직 원본 네트워크의 피드백에만 의존하여 학습하기 때문에, 양자화된 네트워크의 낮은 bit-width로 인한 감소된 수용 능력 차이를 학습에 반영하지 못한다. 제안한 BAG는 양자화된 네트워크와 동일한 bit-width로 양자화하여, 양자화된 네트워크에 맞는 합성 이미지를 생성하여 이러한 문제를 완화한다. 또한, 양자화된 네트워크와 원본 모델 간의 지식 격차를 줄이는 것 역시 양자화에서 매우 중요한 문제이다. 이를 완화하기 위해 제안한 채널 어텐션 기반 중간 레이어 지식 증류는 학생 모델이 교사 모델로부터 어떤 채널에 더 집중해서 학습해야 하는지를 가르친다. 제안한 기법의 효율성을 보이기 위해, CIFAR-100에서 학습한 원본 네트워크를 가중치와 활성값을 각각 3-bit로 양자화하여 학습을 수행하였다. 그 결과 56.14%의 Top-1 Accuracy를 달성하였으며, 베이스라인 모델인 AdaDFQ 대비 3.4% 정확도를 향상했다.

• International Journal of Computer Science & Network Security
• /
• 제24권10호
• /
• pp.206-212
• /
• 2024

In the last 10 years, artificial intelligence (AI) has shown more predictive accuracy than humans in many fields. Its promising future founded on its great performance increases people's concern about its black-box mechanism. In many fields, such as medicine, mistakes lacking explanations are hardly accepted. As a result, research on interpretable AI is of great significance. Although much work about interpretable AI methods are common in classification tasks, little has focused on segmentation tasks. In this paper, we explored the interpretability on a Deep Retinal Image Understanding (DRIU) network, which is used to segment the vessels from retinal images. We combine the Grad Class Activation Mapping (Grad-CAM), commonly used in image classification, to generate saliency map, with the segmentation task network. Through the saliency map, we got information about the contribution of each layer in the network during predicting the vessels. Therefore, we adjusted the weights of last convolutional layer manually to prove the accuracy of the saliency map generated by Grad-CAM. According to the result, we found the layer 'upsample2' to be the most important during segmentation, and we improved the mIoU score (an evaluation method) to some extent.