Fetulhak Abdurahman

Lecturer | Deep Learning Enthusiast

Summary

I am a believer and believe that Artificial Intelligence (AI) will become the next electricity in the world. Its impact will transform the world how electricity transformed the world 100 years ago. Over the past decade AI has been advanced due to the vast availability of digital data (such as medical images and other medical record data, text data, satellite images, etc.), availability of powerful computing resources (such as GPUs, TPUs, cloud platforms, etc.) and the plethora of machine learning and deep learning algorithms (e.g look what the CNNs did on Computer vision (CV) and Transformers did on Natural language processing (NLP).
AI techniques have been applied in various sectors such as health, education, security systems, games (Alpha Go), recommender and dialog systems (Alexa), and many more. Today the scientific community is more concerned about the ethical issues raised by AI-enabled technologies ( e.g face detection algorithms will be biased towards a specific group if they are not trained on all-inclusive demographic data). Today AI algorithms are being used in overly complex and sensitive health/medical sectors (e.g diagnosis and detection of brain tumor, breast cancer detection, chest x-ray analysis, and many more).
When we design and develop AI applications and AI solutions we have to consider the ethical, societal, and economic implications of AI and we have to have a broad range of exposures to the opportunities and challenges we face during the development of AI applications.
My interests include artificial intelligence, machine learning, computer vision, medical image analysis, and natural language processing. I received my MSc degree from Addis Ababa University in Ethiopia, where I took advanced computer engineering courses and work on my thesis by developing a security auditing tool for Android mobile devices. I was introduced to the machine learning world during my MSc thesis, and I was always enthusiastic about it. I am currently working as a Lecturer at Jimma University where I work both on teaching and advising students on their projects and research. I am also working dedicatedly as an independent researcher on medical imaging and natural language processing by applying the techniques of ML and DL.
Further, I'm interested in reading scientific findings, participating in and organizing scientific events ( e.g seminars, reading groups, conferences), and discussing ideas with friends.



My Skills


Few of the skills that I consider myself acquainted with, to serve the given opportunity are showcased here.



Work Experience




Projects




Selected Publications

Automating cytology-based cervical cancer screening could alleviate the shortage of skilled pathologists in developing countries. Up until now, computer vision experts have attempted numerous semi and fully automated approaches to address the need. Yet, these days, leveraging the astonishing accuracy and reproducibility of deep neural networks has become common among computer vision experts. In this regard, the purpose of this study is to classify single-cell Pap smear (cytology) images using pre-trained deep convolutional neural network (DCNN) image classifiers. We have fine-tuned the top ten pre-trained DCNN image classifiers and evaluated them using five class single-cell Pap smear images from SIPaKMeD dataset. The pre-trained DCNN image classifiers were selected from Keras Applications based on their top 1% accuracy. Our experimental result demonstrated that from the selected top-ten pre-trained DCNN image classifiers DenseNet169 outperformed with an average accuracy, precision, recall, and F1-score of 0.990, 0.974, 0.974, and 0.974, respectively. Moreover, it dashed the benchmark accuracy proposed by the creators of the dataset with 3.70%. Even though the size of DenseNet169 is small compared to the experimented pre-trained DCNN image classifiers, yet, it is not suitable for mobile or edge devices. Further experimentation with mobile or small-size DCNN image classifiers is required to extend the applicability of the models in real-world demands. In addition, since all experiments used the SIPaKMeD dataset, additional experiments will be needed using new datasets to enhance the generalizability of the models.
BMC Biomedical Engineering, 2021

Amharic is the ofcial language of the Federal Government of Ethiopia, with more than 27 million speakers. It uses an Ethiopic script, which has 238 core and 27 labialized characters. It is a low-resourced language, and a few attempts have been made so far for its handwritten text recognition. However, Amharic handwritten text recognition is challenging due to the very high similarity between characters. This paper presents a convolutional recurrent neural networks based ofine handwritten Amharic word recognition system. The proposed framework comprises convolutional neural networks (CNNs) for feature extraction from input word images, recurrent neural network (RNNs) for sequence encoding, and connectionist temporal classifcation as a loss function. We designed a custom CNN model and compared its performance with three diferent state-of-the-art CNN models, including DenseNet-121, ResNet-50 and VGG-19 after modifying their architectures to ft our problem domain, for robust feature extraction from handwritten Amharic word images. We have conducted detailed experiments with diferent CNN and RNN architectures, input word image sizes, and applied data augmentation techniques to enhance performance of the proposed models. We have prepared a handwritten Amharic word dataset, HARD-I, which is available publicly for researchers. From the experiments on various recognition models using our dataset, a WER of 5.24 % and CER of 1.15 % were achieved using our best-performing recognition model. The proposed models achieve a competitive performance compared to existing models for ofine handwritten Amharic word recognition.
SN Applied Sciences, 2021

In this paper, we explore the applicability of deep learning based object detection algorithms for malaria parasite detection using microscopic images captured using mobile phone camera We modified one-stage object detector models, YOLOv3 and YOLOv4 models to make them suitable for small object detection problems. We have also done comprehensive comparative analysis with other one-stage and two-stage object detection algorithms. The experimental results of this study demonstrate that performance of modified YOLOV3 and YOLOV4 models are highly promising for detecting malaria parasites from images captured by a smartphone camera over the microscope eyepiece. The proposed system is suitable for deployment in low-resource setting areas.
BMC Bioinformatics, 2021

In this work we proposed a system using convolutional neural networks for recognition of the handwritten Amharic characters. There are two major units in the system: the first one is preprocessing of collected data and segmentation of characters and the second one is extracting learnable features from character images and using those features for classification. Digitalization, noise removal, binarization, normalization and others belongs to the preprocessing step in the proposed system. In the segmentation step different segmentation methods are used such as line segmentation, word segmentation and character segmentation from the given scanned image of handwritten document to extract individual characters. The two essential components in recognition, feature extraction and classification, both are done in our CNN model.
DergiPark, 2019

Image quality assessment methods are used in different image processing applications. Among them, image compression and image super-resolution can be mentioned in wireless capsule endoscopy (WCE) applications. The existing image compression algorithms for WCE employ the generalpurpose image quality assessment (IQA) methods to evaluate the quality of the compressed image. Due to the specific nature of the images captured by WCE, the general-purpose IQA methods are not optimal and give less correlated results to that of subjective IQA (visual perception). This paper presents improved image quality assessment techniques for wireless capsule endoscopy applications. The proposed objective IQA methods are obtained by modifying the existing full-reference image quality assessment techniques. The modification is done by excluding the noninformative regions, in endoscopic images, in the computation of IQA metrics. The experimental results demonstrate that the proposed IQA method gives an improved peak signal-tonoise ratio (PSNR) and structural similarity index (SSIM). The proposed image quality assessment methods are more reliable for compressed endoscopic capsule images.
EEEIC International Publishing, 2020

In this MSc thesis we build a Java based framework that consists of real time monitoring, collection and analysis of various features of the android platform for detection of malicius behaviour.
MSc Manuscript,2014

Recent Publications

. Single-cell conventional pap smear image classification using pre-trained deep neural network architectures. BMC Biomedical Engineering, 2021.

. AHWR-Net: offline handwritten amharic word recognition using convolutional recurrent neural network. SN Applied Sciences, 2021.

. Malaria parasite detection in thick blood smear microscopic images using modified YOLOV3 and YOLOV4 models. BMC Bioinformatics, 2021.

Recent & Upcoming Talks

Artificial Intelligence for Automated Screening of Malaria Disease
Thur, January 20, 2022
Public MSc Defense
Wed, Oct 13, 2014

Teaching

Academic courses

  • Research methodology | 2020 - now

  • Embedded System Design | 2017 - 2018

  • Final Year Project for Undergraduate Students of Jimma University | 2015 - now

    Agro-Robot and Smart Glove | 2019

    3D printing machine Design | 2018

    Quad-copter drone design for agricultural purpose | 2017

    Automated solid waste managment system | 2016

    Transport managment system | 2015

Master thesisses

  • Recurrent Neural Network based Lost Frames Recovery Algorithm for MultiView Video Transmission over Wireless Sensor Network

  • Voice Biometric Based Forensic Speaker Recognition Using Machine Learning

  • Handwritten Text recognition system using deep learning

  • Hate speech detection from social media content using deep learning.

  • Named Entity Recognition.

  • Speech recognition system.

  • Drought prediction system using satellite images.

  • Detection and classification plant parts for traditional medicinal purpose using deep learning.

  • Detection and classification of adult movie/video content using deep learning.

  • Network intrusion detection system using machine learning.