- 标题
- 摘要
- 关键词
- 实验方案
- 产品
-
[IEEE 2018 IEEE 12th International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC) - Hanoi (2018.9.12-2018.9.14)] 2018 IEEE 12th International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC) - Designing Compact Convolutional Neural Network for Embedded Stereo Vision Systems
摘要： Autonomous systems are used in a wide range of domains from indoor utensils to autonomous robot surgeries and self-driving cars. Stereo vision cameras probably are the most flexible sensing way in these systems since they can extract depth, luminance, color, and shape information. However, stereo vision based applications suffer from huge image sizes and computational complexity leading system to higher power consumption. To tackle these challenges, in the first step, GIMME2 stereo vision system [1] is employed. GIMME2 is a high-throughput and cost efficient FPGA-based stereo vision embedded system. In the next step, we present a framework for designing an optimized Deep Convolutional Neural Network (DCNN) for time constraint applications and/or limited resource budget platforms. Our framework tries to automatically generate a highly robust DCNN architecture for image data receiving from stereo vision cameras. Our proposed framework takes advantage of a multi-objective evolutionary optimization approach to design a near-optimal network architecture for both the accuracy and network size objectives. Unlike recent works aiming to generate a highly accurate network, we also considered the network size parameters to build a highly compact architecture. After designing a robust network, our proposed framework maps generated network on a multi/many core heterogeneous System-on-Chip (SoC). In addition, we have integrated our framework to the GIMME2 processing pipeline such that it can also estimate the distance of detected objects. The generated network by our framework offers up to 24x compression rate while losing only 5% accuracy compare to the best result on the CIFAR-10 dataset.
关键词： Deep Convolutional Neural Network,Stereo Vision Systems,Neural Processing Unit,Neural Network Architecture Search
更新于2025-09-23 15:22:29
-
[IEEE 2018 10th International Conference on Modelling, Identification and Control (ICMIC) - Guiyang (2018.7.2-2018.7.4)] 2018 10th International Conference on Modelling, Identification and Control (ICMIC) - Automatic Segmentation and 3D Reconstruction of Spine Based on FCN and Marching Cubes in CT Volumes
摘要： The spine is of great significance in the course of radiotherapy. The accurate location of the spine can provide reference for the determination of the tumor target area and the endanger organ in the radiotherapy plan. However, for some low-resolution areas of CT images, traditional methods cannot achieve a good segmentation effect. Due to the lack of data marked by doctors, there are few studies on the use of deep learning methods for segmentation of the spine. We use threshold segmentation and manual labeling methods to make our own data sets. This article combines the Fully Convolutional Neural Network (FCN) and the Marching Cubes (MC) algorithms to automatically segment and reconstruct the spine in the CT images. And we improved the network structure of FCN because FCN finally lost many details in one step down sampling. In the study, we used data from 40 patients, of which 30 were for training and 10 for testing. The final segmentation accuracy of the improved network is over 93%. The experimental results show that this method has a good segmentation effect and can better restore the shape of the spine and ribs. This preliminary result showed that our spine segmentation method had a great potential to reduce human efforts in labeling CT images in radiation therapy.
关键词： Fully Convolutional Neural Network,Spine,Medical Image,Marching Cubes
更新于2025-09-23 15:22:29
-
[IEEE 2018 2nd International Conference on Data Science and Business Analytics (ICDSBA) - Changsha, China (2018.9.21-2018.9.23)] 2018 2nd International Conference on Data Science and Business Analytics (ICDSBA) - Detection of Diabetic Retinopathy Images Using a Fully Convolutional Neural Network
摘要： The paper discusses the development and application of a convolutional neural network (CNN) model for digital image processing in the context of data science and business analytics. It focuses on improving the accuracy and efficiency of image classification tasks.
关键词： Image Classification,Digital Image Processing,Business Analytics,Data Science,Convolutional Neural Network
更新于2025-09-23 15:22:29
-
[IEEE 2018 IEEE International Conference on Imaging Systems and Techniques (IST) - Krakow (2018.10.16-2018.10.18)] 2018 IEEE International Conference on Imaging Systems and Techniques (IST) - Robust Estimation of Product Amount on Store Shelves from a Surveillance Camera for Improving On-Shelf Availability
摘要： This paper proposes a method to robustly estimate product amount on store shelves from a surveillance camera for improving on-shelf availability. We focus on changes of products on the shelves such as “product taken (decrease)” and “product replenished/returned (increase)”, and compute product amount by accurately accumulating them. The proposed method first detects change regions of products on the shelves in an image using background subtraction followed by moving object removal. The detected change regions are then classified into several classes representing the actual changes on the shelves such as “product taken” by using convolutional neural networks. Finally, the changes of products on the shelves are accumulated using classification results, and product amount on the shelves visible in the image is computed as on-shelf availability. Experimental results using two videos captured in a real store show that our method achieves success rate of 89.6% for on-shelf availability when an error margin is within one product. With high accuracy, store clerks can keep high on-shelf availability, enabling the improvement of business profit in retail stores.
关键词： product amount,image processing,on-shelf availability,convolutional neural network,surveillance camera,retail,background subtraction
更新于2025-09-23 15:22:29
-
Multi-level Features Convolutional Neural Network for Multi-focus Image Fusion
摘要： Multi-focus image fusion is an important technique that aims to generate a single clean image by fusing multiple input images. In this paper, we propose a novel multi-level features convolutional neural network (MLFCNN) architecture for image fusion. In the MLFCNN model, all features learned from previous layers are passed to the subsequent layer. Inside every path between the previous layer and the subsequent layer, we add a 1x1 convolution module to reduce the redundancy. In our method, the source images first are fed to our pre-trained MLFCNN model to obtain the initial focus map. Then, the initial focus map is performed by morphological opening and closing operations and followed by a Gaussian filter to obtain the final decision map. Finally, the fused all-in-focus image is generated based on a weighted-sum strategy with the decision map. The experimental results demonstrate that the proposed method outperforms some state-of-the-art image fusion algorithms in terms of both qualitative and objective evaluations.
关键词： convolutional neural network,decision map,multi-focus image fusion,multi-level features
更新于2025-09-23 15:22:29
-
[IEEE 2018 10th International Conference on Wireless Communications and Signal Processing (WCSP) - Hangzhou (2018.10.18-2018.10.20)] 2018 10th International Conference on Wireless Communications and Signal Processing (WCSP) - Dynamic Hand Gesture Recognition Using FMCW Radar Sensor for Driving Assistance
摘要： Dynamic hand gesture recognition is very important for human-computer interaction. In vehicles, hand gesture recognition can be used as the driver's auxiliary system to achieve remote control of the instrument. To a certain extent, this system can avoid physical buttons and touch screens causing interference to the driver. In this paper, we describe a driver-assisted dynamic gesture recognition system to classify nine hand gestures based on micro-Doppler signatures obtained by 77GHz FMCW radar using a convolutional neural network (CNN). We further explore the changes in the accuracy of same gestures in a variety of experimental scenarios to help optimize the robustness of the system.
关键词： convolutional neural network,hand gesture recognition,driver assistance system,FMCW radar sensor
更新于2025-09-23 15:22:29
-
High Efficient Deep Feature Extraction and Classification of Spectral-Spatial Hyperspectral Image Using Cross Domain Convolutional Neural Networks
摘要： Recently, numerous remote sensing applications highly depend on the hyperspectral image (HSI). HSI classification, as a fundamental issue, has attracted increasing attention and become a hot topic in the remote sensing community. We implemented a regularized convolutional neural network (CNN), which adopted dropout and regularization strategies to address the overfitting problem of limited training samples. Although many kinds of the literature have confirmed that it is an effective way for HSI classification to integrate spectrum with spatial context, the scaling issue is not fully exploited. In this paper, we propose a high efficient deep feature extraction and the classification method for the spectral-spatial HSI, which can make full use of multiscale spatial feature obtained by guided filter. The proposed approach is the first attempt to lean a CNN for spectral and multiscale spatial features. Compared to its counterparts, experimental results show that the proposed method can achieve 3% improvement in accuracy, according to various datasets such as Indian Pines, Pavia University, and Salinas.
关键词： Convolutional neural network (CNN),hyperspectral image (HSI) classification,guided filter,spectral-spatial fusion
更新于2025-09-23 15:22:29
-
Infrared super-resolution imaging using multi-scale saliency and deep wavelet residuals
摘要： Infrared (IR) imaging systems with low-density focal plane arrays produce images with poor spatial resolution. To address this limitation, super-resolution (SR) algorithms can be applied on IR-low resolution (LR) images. In this paper, we present a new SR technique based on the multi-scale saliency detection and the residuals learned by the deep convolutional neural network (CNN) in the wavelet domain (DWCNN). The input LR image is processed in the transformed domain by applying 2D discrete wavelet transform. It decomposes an image into its low-frequency and high-frequency subbands. The multi-scale saliency detection is used to extract small scale and large scale salient feature maps from the bicubic upscaled LR image. These maps are incorporated in the high-frequency subbands of the LR image. Furthermore, the low-frequency and high-frequency subands are re?ned using the residuals learned by the DWCNN in training phase. The proposed algorithm is compared with the conventional and state-of-the-art SR methods. Results indicate that our method yields good reconstruction quality with high peak signal to ratio, structural similarity and low blur indices. Besides, our method requires less computational time.
关键词： Infrared imaging,Convolutional neural network,Discrete wavelet transform,Multi-scale saliency,Super-resolution
更新于2025-09-23 15:22:29
-
Hyperspectral Coastal Wetland Classification Based on a Multiobject Convolutional Neural Network Model and Decision Fusion
摘要： The phenomenon of spectral aliasing exists for coastal wetland object types, which leads to class mixing. This letter proposes a multiobject convolutional neural network (CNN) decision fusion classification method for hyperspectral images of coastal wetlands. This method adopts decision fusion based on fuzzy membership rules applied to single-object CNN classification to obtain higher classification accuracy. Experimental results demonstrate the effectiveness of the proposed method for the six object types, including water, tidal flat, reed, and other vegetation types. The overall accuracy of the decision fusion classification method based on fuzzy membership is 82.11%, which is 3.33% and 6.24% higher than those of single-object feature band CNN and support vector machine methods. The classification method based on multiobject CNN decision fusion inherits the characteristics of single-object feature bands of the CNN, making it a practical approach to image classification under the challenging conditions in which class mixing occurs.
关键词： decision fusion,convolutional neural network (CNN),hyperspectral image,Classification
更新于2025-09-23 15:22:29
-
A Pipeline Neural Network For Low-Light Image Enhancement
摘要： Low-light image enhancement is an important challenge in computer vision. Most of low-light images taken in low-light conditions usually look noisy and dark, which makes it more difficult for subsequent computer vision tasks. In this paper, inspired by multi-scale retinex, we present a low-light image enhancement pipeline network based on an end-to-end fully convolutional networks and discrete wavelet transformation (DWT). Firstly, we show that Multi Scale Retinex (MSR) can be considered as a convolutional neural network (CNN) with Gaussian convolution kernel and blending the result of DWT can improve the image produced by MSR. Secondly, we propose our pipeline neural network, consisting of denoising net and low light image enhancement net (LLIE-net) which learns a function from a pair of dark and bright images. Finally, we evaluate our method both in synthetic dataset and public dataset. Experiments reveal that in comparison with other state-of-the-art methods, our methods achieve better performance in the perspective of qualitative and quantitative analysis.
关键词： Convolutional Neural Network,LLIE-Net,Low-light image enhancement
更新于2025-09-23 15:22:29