Anna Sergeevna Bosman

Anna Sergeevna Bosman

Auto-generated profile, based on publicly available information

Anna Sergeevna Bosman

Department of Computer Science, University of Pretoria, Pretoria, South Africa

Research Focus & Metrics

Academic Career Length

19Years (2008-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications

1 July 2019
Verified
Conference Paper
Closed access
KJKevin Louis de Jong
Kevin Louis de Jong
ABAnna Sergeevna Bosman
Anna Sergeevna Bosman
This paper proposes an efficient unsupervised method for detecting relevant changes between two temporally different images of the same scene. A convolutional neural network (CNN) for semantic segmentation is implemented to extract compressed image features, as well as to classify the detected changes into the correct semantic classes. A difference image is created using the feature map information generated by the CNN, without explicitly training on target difference images. Thus, the proposed change detection method is unsupervised, and can be performed using any CNN model pre-trained for semantic segmentation.

19 July 2024 in Pattern Recognition
Verified
Journal Article
Open access
SPSavvas Panagiotou
Savvas Panagiotou
ABAnna S. Bosman
Anna S. Bosman

7 August 2008 in BMC Bioinformatics
Verified
Journal Article
Open access
PDF file
HGHamilton Ganesan
Hamilton Ganesan
ARAnna S Rakitianskaia
Anna S Rakitianskaia
CDColin Davenport
Colin Davenport
BTBurkhard Tümmler
Burkhard Tümmler
OROleg Reva
Oleg Reva
Background: Data mining in large DNA sequences is a major challenge in microbial genomics and bioinformatics. Oligonucleotide usage (OU) patterns provide a wealth of information for large scale sequence analysis and visualization. The purpose of this research was to make OU statistical analysis available as a novel web-based tool for functional genomics and annotation. The tool is also available as a downloadable package. Results: The SeqWord Genome Browser (SWGB) was developed to visualize the natural compositional variation of DNA sequences. The applet is also used for identification of divergent genomic regions both in annotated sequences of bacterial chromosomes, plasmids, phages and viruses, and in raw DNA sequences prior to annotation by comparing local and global OU patterns. The applet allows fast and reliable identification of clusters of horizontally transferred genomic islands, large multi-domain genes and genes for ribosomal RNA. Within the majority of genomic fragments (also termed genomic core sequence), regions enriched with housekeeping genes, ribosomal proteins and the regions rich in pseudogenes or genetic vestiges may be contrasted. Conclusion: The SWGB applet presents a range of comprehensive OU statistical parameters calculated for a range of bacterial species, plasmids and phages. It is available on the Internet at http://www.bi.up.ac.za/SeqWord/mhhapplet.php.

1 July 2019
Verified
Conference Paper
Closed access
GWGerard Jacques van Wyk
Gerard Jacques van Wyk
ABAnna Sergeevna Bosman
Anna Sergeevna Bosman
Convolutional neural network (CNN) architectures have traditionally been explored by human experts in a manual search process that is time-consuming and ineffectively explores the massive space of potential solutions. Neural architecture search (NAS) methods automatically search the space of neural network hyperparameters in order to find optimal task-specific architectures. NAS methods have discovered CNN architectures that achieve state-of-the-art performance in image classification among other tasks, however the application of NAS to image-to- image regression problems such as image restoration is sparse. This paper proposes a NAS method that performs computation- ally efficient evolutionary search of a minimally constrained net- work architecture search space. The performance of architectures discovered by the proposed method is evaluated on a variety of image restoration tasks applied to the ImageNet64x64 dataset, and compared with human-engineered CNN architectures. The best neural architectures discovered using only 2 GPU-hours of evolutionary search exhibit comparable performance to the human-engineered baseline architecture.

28 November 2022
Verified
Book Chapter
Closed access
PDF file
TMThamsanqa Mlotshwa
Thamsanqa Mlotshwa
HDHeinrich van Deventer
Heinrich van Deventer
ABAnna Sergeevna Bosman
Anna Sergeevna Bosman