scholarly journals Analysis of the ISIC Image Datasets: Usage, Benchmarks and Recommendations

2021 ◽  
pp. 102305
Author(s):  
Bill Cassidy ◽  
Connah Kendrick ◽  
Andrzej Brodzicki ◽  
Joanna Jaworek-Korjakowska ◽  
Moi Hoon Yap
Keyword(s):  
2021 ◽  
Vol 7 (2) ◽  
pp. 37
Author(s):  
Isah Charles Saidu ◽  
Lehel Csató

We present a sample-efficient image segmentation method using active learning, we call it Active Bayesian UNet, or AB-UNet. This is a convolutional neural network using batch normalization and max-pool dropout. The Bayesian setup is achieved by exploiting the probabilistic extension of the dropout mechanism, leading to the possibility to use the uncertainty inherently present in the system. We set up our experiments on various medical image datasets and highlight that with a smaller annotation effort our AB-UNet leads to stable training and better generalization. Added to this, we can efficiently choose from an unlabelled dataset.


Cells ◽  
2021 ◽  
Vol 10 (4) ◽  
pp. 846
Author(s):  
Gabrielle Vieyres

This article targets cell biologists who use fluorescence microscopy but lack automatic tools to summarize and manage their image datasets. When using microscopy to document a phenotype, multiple and random pictures are required to reflect the biological diversity of each imaged sample. Managing, integrating and summarizing the acquired data can be a daunting task that becomes extremely time-consuming unless one automatizes it. Unfortunately, if many biologists use microscopy, only a few have automatized procedures to cope with the data generated. For the majority of microscope users, the two developed complementary ImageJ plugins, PicPreview and PicSummary, will allow, in a few clicks and in an instant, to obtain an overview of all pictures taken for each sample of an experiment and a summary with one user-defined representative picture per sample. The plugins and a video tutorial, as well as demonstration pictures, are available as supplementary data at the journal website. PicPreview and PicSummary should save precious time in managing microscopy datasets and in preparing figures for publications.


Author(s):  
M. A. Dogon-Yaro ◽  
P. Kumar ◽  
A. Abdul Rahman ◽  
G. Buyuksalih

Mapping of trees plays an important role in modern urban spatial data management, as many benefits and applications inherit from this detailed up-to-date data sources. Timely and accurate acquisition of information on the condition of urban trees serves as a tool for decision makers to better appreciate urban ecosystems and their numerous values which are critical to building up strategies for sustainable development. The conventional techniques used for extracting trees include ground surveying and interpretation of the aerial photography. However, these techniques are associated with some constraints, such as labour intensive field work and a lot of financial requirement which can be overcome by means of integrated LiDAR and digital image datasets. Compared to predominant studies on trees extraction mainly in purely forested areas, this study concentrates on urban areas, which have a high structural complexity with a multitude of different objects. This paper presented a workflow about semi-automated approach for extracting urban trees from integrated processing of airborne based LiDAR point cloud and multispectral digital image datasets over Istanbul city of Turkey. The paper reveals that the integrated datasets is a suitable technology and viable source of information for urban trees management. As a conclusion, therefore, the extracted information provides a snapshot about location, composition and extent of trees in the study area useful to city planners and other decision makers in order to understand how much canopy cover exists, identify new planting, removal, or reforestation opportunities and what locations have the greatest need or potential to maximize benefits of return on investment. It can also help track trends or changes to the urban trees over time and inform future management decisions.


2020 ◽  
Vol 34 (06) ◽  
pp. 10243-10250
Author(s):  
Jozsef Nemeth

We consider the disentanglement of the representations of the relevant attributes of the data (content) from all other factors of variations (style) using Variational Autoencoders. Some recent works addressed this problem by utilizing grouped observations, where the content attributes are assumed to be common within each group, while there is no any supervised information on the style factors. In many cases, however, these methods fail to prevent the models from using the style variables to encode content related features as well. This work supplements these algorithms with a method that eliminates the content information in the style representations. For that purpose the training objective is augmented to minimize an appropriately defined mutual information term in an adversarial way. Experimental results and comparisons on image datasets show that the resulting method can efficiently separate the content and style related attributes and generalizes to unseen data.


2010 ◽  
Vol 21 (3) ◽  
pp. 644-652 ◽  
Author(s):  
Anja Apel ◽  
Joel G. Fletcher ◽  
Jeff L. Fidler ◽  
David M. Hough ◽  
Lifeng Yu ◽  
...  

Entropy ◽  
2019 ◽  
Vol 21 (11) ◽  
pp. 1062 ◽  
Author(s):  
Yuhang Dong ◽  
W. David Pan ◽  
Dongsheng Wu

Malaria is a severe public health problem worldwide, with some developing countries being most affected. Reliable remote diagnosis of malaria infection will benefit from efficient compression of high-resolution microscopic images. This paper addresses a lossless compression of malaria-infected red blood cell images using deep learning. Specifically, we investigate a practical approach where images are first classified before being compressed using stacked autoencoders. We provide probabilistic analysis on the impact of misclassification rates on compression performance in terms of the information-theoretic measure of entropy. We then use malaria infection image datasets to evaluate the relations between misclassification rates and actually obtainable compressed bit rates using Golomb–Rice codes. Simulation results show that the joint pattern classification/compression method provides more efficient compression than several mainstream lossless compression techniques, such as JPEG2000, JPEG-LS, CALIC, and WebP, by exploiting common features extracted by deep learning on large datasets. This study provides new insight into the interplay between classification accuracy and compression bitrates. The proposed compression method can find useful telemedicine applications where efficient storage and rapid transfer of large image datasets is desirable.


Sign in / Sign up

Export Citation Format

Share Document