scholarly journals Semantic segmentation of slums in satellite images using transfer learning on fully convolutional neural networks

2019 ◽  
Vol 150 ◽  
pp. 59-69 ◽  
Author(s):  
Michael Wurm ◽  
Thomas Stark ◽  
Xiao Xiang Zhu ◽  
Matthias Weigand ◽  
Hannes Taubenböck
Keyword(s):  
Neural Networks ◽  
Transfer Learning ◽  
Convolutional Neural Networks ◽  
Satellite Images ◽  
Semantic Segmentation ◽  
Fully Convolutional Neural Networks

Breast Infrared Thermography Segmentation Based on Adaptive Tuning of a Fully Convolutional Network

2020 ◽  
Vol 16 (5) ◽  
pp. 611-621
Author(s):  
Mazhar Basyouni Tayel ◽  
Azza Mahmoud Elbagoury
Keyword(s):  
Neural Networks ◽  
Infrared Thermography ◽  
Transfer Learning ◽  
Convolutional Neural Networks ◽  
Data Augmentation ◽  
Semantic Segmentation ◽  
Image Features ◽  
Fine Tuning ◽  
Fully Convolutional Neural Networks ◽  
End To End

Background: Accurate segmentation of Breast Infrared Thermography is an important step for early detection of breast pathological changes. Automatic segmentation of Breast Infrared Thermography is a very challenging task, as it is difficult to find an accurate breast contour and extract regions of interest from it. Although several semi-automatic methods have been proposed for segmentation, their performance often depends on hand-crafted image features, as well as preprocessing operations. Objective: In this work, an approach to automatic semantic segmentation of the Breast Infrared Thermography is proposed based on end-to-end fully convolutional neural networks and without any pre or post-processing. Methods: The lack of labeled Breast Infrared Thermography data limits the complete utilization of fully convolutional neural networks. The proposed model overcomes this challenge by applying data augmentation and two-tier transfer learning from bigger datasets combined with adaptive multi-tier fine-tuning before training the fully convolutional neural networks model. Results: Experimental results show that the proposed approach achieves better segmentation results: 97.986% accuracy; 98.36% sensitivity and 97.61% specificity compared to hand-crafted segmentation methods. Conclusion: This work provided an end-to-end automatic semantic segmentation of Breast Infrared Thermography combined with fully convolutional networks, adaptive multi-tier fine-tuning and transfer learning. Also, this work was able to deal with challenges in applying convolutional neural networks on such data and achieving the state-of-the-art accuracy.

scholarly journals Exact Tile-Based Segmentation Inference for Images Larger than GPU Memory

2021 ◽  
Vol 126 ◽  
Author(s):  
Michael Majurski ◽  
Peter Bajcsy
Keyword(s):  
Neural Networks ◽  
Convolutional Neural Networks ◽  
Semantic Segmentation ◽  
Tile Size ◽  
Fully Convolutional Neural Networks ◽  
Memory Size ◽  
Slide Scanner ◽  
Microscopy Images

We address the problem of performing exact (tiling-error free) out-of-core semantic segmentation inference of arbitrarily large images using fully convolutional neural networks (FCN). FCN models have the property that once a model is trained, it can be applied on arbitrarily sized images, although it is still constrained by the available GPU memory. This work is motivated by overcoming the GPU memory size constraint without numerically impacting the fnal result. Our approach is to select a tile size that will ft into GPU memory with a halo border of half the network receptive feld. Next, stride across the image by that tile size without the halo. The input tile halos will overlap, while the output tiles join exactly at the seams. Such an approach enables inference to be performed on whole slide microscopy images, such as those generated by a slide scanner. The novelty of this work is in documenting the formulas for determining tile size and stride and then validating them on U-Net and FC-DenseNet architectures. In addition, we quantify the errors due to tiling confgurations which do not satisfy the constraints, and we explore the use of architecture effective receptive felds to estimate the tiling parameters.

scholarly journals EXTRACTION OF BUILT-UP AREAS USING CONVOLUTIONAL NEURAL NETWORKS AND TRANSFER LEARNING FROM SENTINEL-2 SATELLITE IMAGES

2018 ◽  
Vol XLII-3 ◽  
pp. 79-85 ◽  
Author(s):  
V. S. Bramhe ◽  
S. K. Ghosh ◽  
P. K. Garg
Keyword(s):  
Neural Networks ◽  
Transfer Learning ◽  
Convolutional Neural Networks ◽  
Large Scale ◽  
Satellite Images ◽  
Spatial Information ◽  
Back Propagation ◽  
Back Propagation Neural Network ◽  
Support Vector ◽  
Sentinel 2

With rapid globalization, the extent of built-up areas is continuously increasing. Extraction of features for classifying built-up areas that are more robust and abstract is a leading research topic from past many years. Although, various studies have been carried out where spatial information along with spectral features has been utilized to enhance the accuracy of classification. Still, these feature extraction techniques require a large number of user-specific parameters and generally application specific. On the other hand, recently introduced Deep Learning (DL) techniques requires less number of parameters to represent more abstract aspects of the data without any manual effort. Since, it is difficult to acquire high-resolution datasets for applications that require large scale monitoring of areas. Therefore, in this study Sentinel-2 image has been used for built-up areas extraction. In this work, pre-trained Convolutional Neural Networks (ConvNets) i.e. Inception v3 and VGGNet are employed for transfer learning. Since these networks are trained on generic images of ImageNet dataset which are having very different characteristics from satellite images. Therefore, weights of networks are fine-tuned using data derived from Sentinel-2 images. To compare the accuracies with existing shallow networks, two state of art classifiers i.e. Gaussian Support Vector Machine (SVM) and Back-Propagation Neural Network (BP-NN) are also implemented. Both SVM and BP-NN gives 84.31 % and 82.86 % overall accuracies respectively. Inception-v3 and VGGNet gives 89.43 % of overall accuracy using fine-tuned VGGNet and 92.10 % when using Inception-v3. The results indicate high accuracy of proposed fine-tuned ConvNets on a 4-channel Sentinel-2 dataset for built-up area extraction.

Sign in / Sign up

Export Citation Format

Share Document