Detection of Diabetic Retinopathy using Convolutional Neural Network

Diabetic Retinopathy is a medical condition in which damage occurs to the retina due to diabetes mellitus. The diagnosis of Diabetic Retinopathy through colored fundus images stand in need of experienced clinicians to identify the presence and significance of many small features, which makes it a time consuming task. In this paper, we propose a CNN based approach to detect Diabetic Retinopathy in fundus images. Data used to train the model is prepocessed by a new segmentation technique using Gabor filters. Due to small dataset, data augmentation is done to get enough data to train the model. Our segmentation model detects intricate features in the fundus images and detect the presence of DR. A high-end Graphics Processor Unit (GPU) is used to train the model efficiently. The publicly available Kaggle Dataset is used to demonstrate impressive results, particularly for a high-level classification task. On the training dataset of 14,650 images, our proposed CNN achieves a specificity of 94% and an accuracy of 69% on 3,660 validation images.

Analysis of GPR Wave Propagation in Complex Underground Structures Using CUDA-Implemented Conformal FDTD Method

Ground penetrating radar (GPR), as a kind of fast, effective, and nondestructive tool, has been widely applied to nondestructive testing of road quality. The finite-difference time-domain method (FDTD) is the common numerical method studying the GPR wave propagation law in layered structure. However, the numerical accuracy and computational efficiency are not high because of the Courant-Friedrichs-Lewy (CFL) stability condition. In order to improve the accuracy and efficiency of FDTD simulation model, a parallel conformal FDTD algorithm based on graphics processor unit (GPU) acceleration technology and surface conformal technique was developed. The numerical simulation results showed that CUDA-implemented conformal FDTD method could greatly reduce computational time and the pseudo-waves generated by the ladder approximation. And the efficiency and accuracy of the proposed method are higher than the traditional FDTD method in simulating GPR wave propagation in two-dimensional (2D) complex underground structures.

A Graphics Processor Unit-Accelerated Freeform Surface Offsetting Method for High-Resolution Subtractive Three-Dimensional Printing (Machining)

Machining is one of the major manufacturing methods having very wide applications in industries. Unlike layer-by-layer additive three-dimensional (3D) printing technology, the lack of an easy and intuitive programmability in conventional toolpath planning approach in machining leads to significantly higher manufacturing cost for direct computer numerical control (CNC)-based prototyping (i.e., subtractive 3D printing). In standard computer-aided manufacturing (CAM) packages, general use of B-rep (boundary representation) and non-uniform rational basis spline (NURBS)-based representations of the computer-aided design (CAD) interfaces make core computations of tool trajectories generation process, such as surface offsetting, difficult. In this work, the problem of efficient generation of freeform surface offsets is addressed with a novel volumetric (voxel) representation. It presents an image filter-based offsetting algorithm, which leverages the parallel computing engines on modern graphics processor unit (GPU). The compact voxel data representation and the proposed computational acceleration on GPU together are capable to process voxel offsetting at four-fold higher resolution in interactive CAM application. Additionally, in order to further accelerate the offset computation, the problem of offsetting with a large distance is decomposed into successive offsetting using smaller distances. The performance trade-offs between accuracy and computation time of the offset algorithms are thoroughly analyzed. The developed GPU implementation of the offsetting algorithm is found to be robust in computation, and demonstrates a 50-fold speedup on single graphics card (NVIDIA GTX780Ti) relative to prior best-performing algorithms developed for multicores central processing units (CPU). The proposed offsetting approach has been validated for a variety of complex parts produced on different multi-axis CNC machine tools including turning, milling, and compound turning-milling.

Efficiency Analysis of Genetic Algorithm and Genetic Programming in Data Mining and Image Processing

Today, in the age of artificial intelligence and machine learning, Data mining and Image processing are two important platforms. GA and GP are value based and program based randomized searching tools respectively and these two are very much useful in the fields' data mining and image processing for handling different issues. In this chapter, a review is made on ability of GA and GP in some applications of these two fields. Here, the selected subfields of data mining are market analysis, fraud detection, risk management, sports analysis, protein interaction, classification of data, drug discovery and feature construction. The similar in image processing are enhancement and segmentation of images, face recognition, photo mosaic generation, data embedding, image pattern classification, object detection and Graphics Processor Unit (GPU) development. The efficiencies of GA and GP in these particular applications are analyzed with corresponding parameters, comparing with other non-GA and non-GP approaches of the corresponding subfields.

Efficiency Analysis of Genetic Algorithm and Genetic Programming in Data Mining and Image Processing

Today, in the age of artificial intelligence and machine learning, Data mining and Image processing are two important platforms. GA and GP are value based and program based randomized searching tools respectively and these two are very much useful in the fields' data mining and image processing for handling different issues. In this chapter, a review is made on ability of GA and GP in some applications of these two fields. Here, the selected subfields of data mining are market analysis, fraud detection, risk management, sports analysis, protein interaction, classification of data, drug discovery and feature construction. The similar in image processing are enhancement and segmentation of images, face recognition, photo mosaic generation, data embedding, image pattern classification, object detection and Graphics Processor Unit (GPU) development. The efficiencies of GA and GP in these particular applications are analyzed with corresponding parameters, comparing with other non-GA and non-GP approaches of the corresponding subfields.

A Graphical Approach for Freeform Surface Offsetting With GPU Acceleration for Subtractive 3D Printing

The lack of plug-and-play programmability in conventional toolpath planning approach in subtractive manufacturing, i.e., machining leads to significantly higher manufacturing cost for CNC based prototyping. In computer aided manufacturing (CAM) packages, typical B-rep or NURBS based representations of the CAD interfaces challenge core computations of tool trajectories generation process, such as, surface offsetting to be completely automated. In this work, the problem of efficient generation of free-form surface offsets is addressed with a novel volumetric representation. It presents an image filter based offsetting algorithm, which leverages the parallel computing engines on modern graphics processor unit (GPU). The scalable voxel data structure and the proposed hardware-accelerated volumetric offsetting together advance the computation and memory efficiencies well beyond the capability of past studies. Additionally, in order to further accelerate the offset computation the problem of offsetting with a large distance is decomposed into successive offsetting using smaller distances. The accuracy of the offset algorithms is thoroughly analyzed. The developed GPU implementation of the offsetting algorithm is robust in computation, easy to comprehend, and achieves a 50-fold speedup on single graphics card (NVIDIA GTX780Ti) relative to prior best-performing dual socket quad-core CPU implementation.