Many researchers have realized the intelligent medical diagnosis of diabetic retinopathy (DR) from fundus images by using deep learning methods, including supervised contrastive learning (SupCon). However, although SupCon brings label information into the calculation of contrastive learning, it does not distinguish between augmented positives and same-label positives. As a result, we propose the concept of Angular Margin and incorporate it into SupCon to address this issue. To demonstrate the effectiveness of our strategy, we tested it on two datasets for the detection and grading of DR. To align with previous work, Accuracy, Precision, Recall, F1, and AUC were selected as evaluation metrics. Moreover, we also chose alignment and uniformity to verify the effect of representation learning and UMAP (Uniform Manifold Approximation and Projection) to visualize fundus image embeddings. In summary, DR detection achieved state-of-the-art results across all metrics, with Accuracy = 98.91, Precision = 98.93, Recall = 98.90, F1 = 98.91, and AUC = 99.80. The grading also attained state-of-the-art results in terms of Accuracy and AUC, which were 85.61 and 93.97, respectively. The experimental results demonstrate that Angular Margin is an excellent intelligent medical diagnostic algorithm, performing well in both DR detection and grading tasks.
In hyperspectral remote sensing, achieving high spatial resolution holds paramount importance for an array of applications, such as environmental monitoring, geographic mapping, and precision agriculture. Nevertheless, conventional hyperspectral images frequently grapple with the issue of restricted spatial resolution. We apply optimized inversion methods to hyperspectral image fusion and present an innovative approach for hyperspectral image fusion which combines the Hue–Intensity–Saturation (HIS) transform, the wavelet transform, and the Trust-Region Conjugate Gradient technique. This amalgamation not only refines spatial precision but also augments spectral faithfulness, which is a pivotal aspect for applications like precise object detection and classification. In the context of our investigation, we conducted a thorough validation of our proposed HIS, Wavelet, and Trust-Region Conjugate Gradient (TRCG-HW) method for image fusion using a comprehensive suite of evaluation metrics. These metrics encompassed the Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), Correlation Coefficient (CC), Spectral Angle Mapper (SAM), and Error Relative Global Accuracy Score (ERGAS). The findings incontrovertibly establish TRCG-HW as the preeminent method among those considered. Our study effectively tackles the pressing predicament of low spatial resolution encountered in hyperspectral imaging. This innovative paradigm harbors the potential to revolutionize high-resolution hyperspectral data acquisition, propelling the field of hyperspectral remote sensing forward and efficiently catering to crucial application.
Currently, computer-generated holograms (CGHs) based on ray tracing technology are generated faster and faster, and the reconstructed scenes are getting bigger and bigger and contain more and more information. Based on this situation, there are also more applications of using CGHs to hide information, but there is a lack of research on the ability to hide information. To address this issue, this paper proposes a point-sampling CGH method based on ray tracing. Our method utilizes ray tracing techniques to rapidly sample text information at different depths in the scene and hides the depth-encoded text information in the carrier image using discrete cosine transform. The reconstructed image after embedding shows good results, with a peak signal-to-noise ratio (PSNR) of 29.56 dB between the hidden images before and after embedding. The PSNR value between the embedded carrier image and the original carrier image is 51.66 dB, making it difficult for the human eye to distinguish, thereby effectively protecting the generated CGH. We also analyzed the maximum information density and observed that computational holograms obtain the maximum information density at ${200} \times {200}$ resolution.
Fast holographic generation technology depends on the swift creation of computer-generated holograms (CGHs). Currently, the fast generation method of computer-generated holograms based on ray tracing is still based on non-physical ray tracing methods, making it difficult to accurately simulate and generate computer-generated holograms. There is a scarcity of research on employing path tracing to swiftly generate CGH with a minimal computational workload and reduced computation time. To address this issue, we employed physically-based path tracing technique for realistic rendering and made improvements on the bidirectional scattering distribution function (BSDF) and ray sampling methods. We have developed a path tracing technique based on improved BSDF and single ray sampling to generate CGHs with realism. In comparison to conventional point cloud and ray tracing methods, the path tracing approach employs distinct shading rendering and lighting models to enhance a scene's sense of realism and generate CGHs rapidly. Our proposed technique takes approximately 10 ms to generate CGHs in various scenes, and the reconstructed images featuring photorealistic appearances are produced through optical reconstruction experiments. This novel method offers an efficient solution for generating CGHs and we attempted to generate a sequence using our method, applying our method to rapidly generate CGH videos.
1 May 2023 in Journal of Physics: Conference Series
The oil level of the power transformer is an important predictor to analyze the working state. The conventional oil level detection method cannot detect the level when the power transformer is working, which is not convenient for the actual application. In this paper, an oil level detection method based on double pressure sensors is proposed to online measure the oil level for power transformer, the basic principle of the proposed oil level detection method is explained and studied, and a prototype is developed to verify the proposed method. The tested results show that the developed prototype can reach high precision for oil level detection of a transformer.
As a special mode of translation, caption translation plays an increasingly important role in the field of translation. Therefore, under the background of international exchanges and cooperation among colleges and universities, it is necessary to deepen the understanding of subtitle translation with the help of relevant theoretical studies. This paper aims to analyze subtitle translation from the perspective of relevance theory. The study of film subtitle translation strategy is of great value not only in communication but also in linguistics. Subtitle translation is a great bridge to communicate international culture and promote the continuous prosperity and development of human culture. This paper takes the subtitle translation of Titanic as an example to describe the overall divided into four parts. It demonstrates whether the subtitle translation of the English movie Titanic is successful or not on the basis of the theoretical discussions and the case study.