Search

Glaucoma is a progressive eye condition that can lead to permanent vision loss. Therefore, on-time detection of glaucoma is critical for making an effective treatment plan. In recent years, enormous attempts have been made to develop automated glaucoma classification systems using CNNs through images. In contrast, limited methods have been proposed for diagnosing different glaucoma stages. It is mainly owing to the lack of large publicly available labeled datasets. Also, fundus images exhibit a high inter-stage resemblance, redundant features and minute size variations of lesions, making the conventional CNNs difficult to classify multiple stages of glaucoma accurately. To address these challenges, this paper proposes a novel adapter and enhanced self-attention based CNN framework named AES-Net for effective classification of glaucoma stages. In particular, we propose a spatial adapter module on top of the backbone network for learning better feature representations and an enhanced self-attention module (ESAM) to capture global feature correlations among the relevant channels and spatial positions. The ESAM assists in capturing stage-specific and detailed-lesion features from the fundus images. Extensive experiments on two multi-stage glaucoma datasets indicate that our AES-Net surpasses CNN-based existing approaches. The Grad-CAM ++ visualization maps further confirm the effectiveness of our AES-Net. • A spatial adapter and enhanced self-attention based CNN, AES-Net, is proposed. • An enhanced self-attention module is proposed to learn class-specific features. • A comparative analysis with state-of-the-art attention mechanisms is performed. • Results on two datasets indicate the effectiveness of AES-Net over current methods. • The visualization results demonstrate the explainability and interpretability of the model. [ABSTRACT FROM AUTHOR]

Image editing technology leads to tampering traces hidden in images hard to be identified, which results in great challenges for image manipulation detection. Therefore, it is necessary to enhance the characteristic response of tampering traces. In this paper, a two-stream network is exploited to extract tampering features from both semantic level and noise level through end-to-end training. At the input end of the network, a multi-scale high-frequency constrained convolution kernel module is proposed to extract multi-channel and multi-scale noise fingerprint information, which enhances the tampering traces. Moreover, we leverage multi-task learning to realize the classification of tampering techniques, the detection and segmentation of tampering regions. The tampered segmentation mask attention is introduced as a novel constraint to be fed back to the detection branch, which further enhances the tampering characteristics and improves the detection performance. Experiments on COVER, CASIA and NIST16 datasets prove that the proposed framework outperforms existing methods quantitatively and qualitatively. The proposed method effectively enhances the characteristics of tampered regions. Moreover, the proposed method has high robustness on JPEG recompression and image resize. • We proposed a multi-scale high-frequency constrained convolution kernel module to extract fingerprint noise information end-to-end, which exploits the domain knowledge of camera model identification. The module expands the receptive field and highlights the contrast of tampering traces at different scales and different channels, which emphasizes the characteristic representation of the tampering traces. • The segmentation mask attention is introduced as a novel constraint to be fed back to the detection branch. The network is forced to pay more attention to the tampered area and enhances the tampering characteristics. The visualization results show that segmentation features can focus on the tampered areas. • The tasks of classification, detection and segmentation are combined as multi-task learning, sharing the characteristics of different tasks to form constraints. The detection performance is further improved. • Extensive experiments on three standard datasets demonstrate that our method outperforms other methods significantly. Moreover, the proposed method has high robustness on JPEG recompression and image resize. [ABSTRACT FROM AUTHOR]

Years ago, data visualizations were almost exclusively used for decision making and data analysis. However, recently visualizations are more utilized for marketing sales tools and many other tasks; and therefore more memorable visualizations are desirable. Thus, studying memorability and automatically assessing visualizations or images memorability is becoming a challenging topic in computer vision. This topic purpose is to list and analyze the characteristics of memorable visualizations/images, in order to establish a predictive model for memorability. Previous methods mainly use techniques to analyze visualizations memorability. In this paper, we use data related to visual attention combined with image features to significantly study visualizations memorability and construct a representative model of visualization memorability. We test our proposed method on MASSVIS data set, by using different classification methods and we significantly reach performed classification rates. [ABSTRACT FROM AUTHOR]

Isomap aims to recover the intrinsic geometric structure of manifold by preserving geodesic distances between all pairs of data points. However it is an unsupervised dimensionality-reduction method. Usually, using class label information can increase the discriminating capability, hence a new supervised Isomap is proposed in this paper, dubbed Multi-manifold Discriminant Isomap (MMD-Isomap). First, data points are partitioned into different manifolds according to their class label information. Then, MMD-Isomap aims at seeking an optimal nonlinear subspace to preserve the geometrical structure of each manifold according to the Isomap criterion, meanwhile, to enhance the discriminating capability by maximizing the distances between data points of different manifolds. Finally, the corresponding optimization problem is solved by using a majorization algorithm. Furthermore, two new numerical metrics are designed to measure the performance of dimensionality-reduction method. In both visualization and classification experiments, MMD-Isomap achieves improved performance over many state-of-the-art methods. [ABSTRACT FROM AUTHOR]