China
Africa
Explore chapters and articles related to this topic
Survey of fabric defect detection using UNet architecture
Published in Sangeeta Jadhav, Rahul Desai, Ashwini Sapkal, Application of Communication Computational Intelligence and Learning, 2022
Saiqa Khan, Ansari Almas Eram, Ansari Saadan, Ansari Raheen Bano, Narjis Fatema Suterwala
The name U-Net is attained from the architecture, which resembles the letter U when viewed, as illustrated in Figure 10.2. A segmented output map is created from the input images. There is no fully connected layer in the network. The convolution layers are the only ones that are used. The original U-Net model is a supervised machine learning method that can effectively segment biological cell images during the testing phase after being fed as many labelled examples as possible during the training phase. U-Net takes a single channel grayscale image and categorises it into two channels depending on the pixel level. Convolutional layers, max-pooling layers, and up-convolutional layers make up this system. Cropping is also required to account for the border pixel difference between every two concatenated convolutions. The cross-entropy loss function of the U-loss net is utilised in biomedical segmentation.
High-Performance Computing and Its Requirements in Deep Learning
Published in Sanjay Saxena, Sudip Paul, High-Performance Medical Image Processing, 2022
Biswajit Jena, Gopal Krishna Nayak, Sanjay Saxena
Inception architecture is known to be one of the most creative architectures in computer vision as its applications range from image recognition to object detection and proved its impact in natural language processing applications, mainly in the design of transformer machines classification tasks, etc. Bert model is a known name in many natural language processing-related fields, which proved its importance mainly in tasks such as text processing, text cleaning, word embedding, machine translation, and other transformer-related applications. It is famous for its state of the art results and ease of use in various applications, and not to forget XL Net, an equally known natural language processing tool. Other important models are Alexnet, VGG16, Mobilenet, etc., which are computer vision-related pre-trained models but are also effective in interdisciplinary research. With this one, we can get the idea of what a pre-trained model is and how effective it can be when provided an opportunity for cross-platform knowledge. U-Net, Segnet and DeconvoNet are three important pre-trained models of image segmentation. U-Net was basically developed for biomedical image segmentation and hence good for all medical images. SegNet is an encoder-decoder-based segmentation technique that has its own advantages and is good for basically natural image segmentation. DeconvoNet follows the same architecture as SegNet however, there are fully connected layers which makes the model larger.
Deep Learning in Brain Segmentation
Published in Saravanan Krishnan, Ramesh Kesavan, B. Surendiran, G. S. Mahalakshmi, Handbook of Artificial Intelligence in Biomedical Engineering, 2021
U-net is a CNN-based neural network introduced by Olaf Ronneberger, Philipp Fischer, and Thomas Brox and was given the name due to its “U”-shaped architecture (Ronneberger et al., 2015). It was originally designed to perform semantic segmentation on biomedical images and is now widely adapted for general segmentation task U-net is an example of the fully convolutional network (FCN) where no fully connected layers are involved. A typical U-net consists of a downsampling path and an upsampling path with a concatenation of the feature maps in between the two paths. The downsampling path contains stacked convolutional layers, activation functions, and pooling layers until it reaches the bottleneck section of the network. With the image gradually passed down the downsampling path, the spatial resolution of the feature map decreases while the feature resolution increases. After another convolution block, the feature maps are passed onto the upsampling path. The upsampling path consists of transpose convolution blocks that up-samples the feature map back to its original dimensions.
Skin disease migration segmentation network based on multi-scale channel attention
Published in Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization, 2023
Bin Yu, Long Yu, Shengwei Tian, Weidong Wu, Zhang Dezhi, Xiaojing Kang
Encoding and decoding networks need to go through two steps: 1) encoding features (image downsampling); 2) decoding features (image upsampling). The upsampling operation makes the edge information of the segmentation result more refined. A typical encoding and decoding structure is U-Net (Ronneberger et al. 2015). In recent years, Biomedical image segmentation has achieved high results, and U-Net has fast training speed, less memory, and a lot of room for improvement. Recently, a variety of U-Net variant network structures have been proposed. LCA-Net has made new breakthroughs in improving the traditional U-Net network. Since it retains the traditional advantages of the U-Net network and does not improve the decoder part, our work is carried out accordingly, See section 3 for details.
Virtual characterisation of porcupine quills using X-ray micro-CT
Published in Virtual and Physical Prototyping, 2023
Yun Lu Tee, Jay R. Black, Phuong Tran
The third segmentation method was applied via the deep learning tool of Dragonfly. The U-net convolutional neural network was chosen for semantic segmentation. U-net is a convolutional network architecture for fast and precise segmentation of images which was first introduced in 2015 for biomedical image segmentation (Ronneberger, Fischer, and Brox 2015). The built-in deep learning tool within the ORS Dragonfly software deep learning segmentation with Segmentation Wizard is used to train and predict the quill and air phase. Five images were initially selected from the CT slices for Otsu thresholding, followed by manual editing the results before using them as a training set. U-net with semantic segmentation was selected to train the data based on manually segmented image slices (Makovetsky, Piche, and Marsh 2018; Tung et al. 2022). The model architecture used was U-net with a depth level of 4 layers. The model was trained with 100 epochs with a batch size of 16, stride ratio of 1.0 with an initial learning rate of 1.0 with the Adadelta optimisation algorithm. The deep learning segmentation was performed in Dragonfly software.
Lane detection based on IBN deep neural network and attention
Published in Connection Science, 2022
Yue Song, Li-yong Wang, Hao-dong Wang, Meng-lin Li
Inspired by the literature (Huang et al., 2019), different types of encoder and decoder combinations can increase the computational efficiency on the premise of ensuring the accuracy of the reference frame. The main structure of the U-Net (Ronneberger et al., 2015) network consists of an encoder, decoder, and bottleneck layer, in which the encoding and decoding are symmetric and skip connection is used. In our study, ResNet34 is used as the encoder module, which gives up average pooling and a fully connected layer retaining the first four feature extraction blocks. The main reason for this is that a shortcut mechanism is introduced by using ResNet34. This avoids gradient disappearance and accelerates network convergence. In the encoder phase, IBN networks are added to ResNet34 to improve network performance without additional computation. The attention mechanism is introduced at the connection between the encoding stage and decoding stage to refine the result of deconvolution by jump connection. The feature map of the attention layer in the encoder is fused with the result of the corresponding deconvolution layer, the structure of which is shown in Figure 1.