Kumar Abhishek
kabhisadasflkjhhe [at] sdfhsdjaffu [dot] 1432@#$2 ca

I am a PhD student in the School of Computing Science at Simon Fraser University, where I work as a part of the Medical Image Analysis Lab (MIAL), under the supervision of Professor Ghassan Hamarneh.

I defended my MSc Thesis in April 2020 on input space augmentation strategies for skin lesion segmentation under the supervision of Professor Ghassan Hamarneh. My examination committee consisted of Professors Mark S. Drew, Sandra Avila, and Angel X. Chang, and my thesis was accepted without any revisions. Previously, I graduated with a Bachelor of Technology in Electronics and Communication Engineering with a focus on Image Processing and Machine Learning from the Indian Institute of Technology (IIT) Guwahati in 2015. My undergraduate thesis was advised by Professor Prithwijit Guha.

During my undergraduate years, I carried out internships at LFOVIA, IIT Hyderabad and CTO Office, Wipro. After graduating from IIT Guwahati, I have worked at Wipro Analytics and Altisource Labs.

CV  |  Google Scholar  |  LinkedIn  |  Blog


I'm interested in computer vision, machine learning, and image processing. At MIAL, I work on applying deep learning methods to medical image analysis. The primary focus of my work has been on adversarial attacks and generative models.


Deep Semantic Segmentation of Natural and Medical Images: A Review
Kumar Abhishek*, Saeid Asgari Taghanaki*, Joseph Paul Cohen, Julien Cohen-Adad, Ghassan Hamarneh [*: Joint first authors]
Artificial Intelligence Review, 2020

We present a comprehensive survey of advances in deep learning-based semantic segmentation of natural and medical images, categorizing the contributions in 6 broad categories, and discuss limitations and potential research directions.


Illumination-based Transformations Improve Skin Lesion Segmentation in Dermoscopic Images
Kumar Abhishek, Ghassan Hamarneh, Mark S. Drew
ISIC Skin Image Analysis Workshop, IEEE International Conference on Computer Vision and Pattern Recognition (CVPR), 2020

We incorporate information from specific color bands, illumination invariant grayscale images, and shading-attenuated images obtained from RGB dermoscopic images of skin lesions to improve the lesion segmentation.



Artificial Intelligence In Glioma Imaging: Challenges and Advances
Weina Jin, Mostafa Fatehi, Kumar Abhishek, Mayur Mallya, Brian Toyota, Ghassan Hamarneh
Journal of Neural Engineering, 2020

We review the literature to analyze the most important challenges in the clinical adoption of AI-based methods and present a summary of the recent advances, categorizing them into three broad categories: dealing with limited data volume and annotations, training of deep learning-based models, and the clinical deployment of these models.


Signed Input Regularization
Kumar Abhishek*, Saeid Asgari Taghanaki*, Ghassan Hamarneh [*: Joint first authors]
arXiv pre-print, arXiv:1911.07086

We propose a new regularization technique which learns to estimate the contribution of the input variables in the final prediction output and can be used as a data augmentation strategy.


Mask2Lesion: Mask-Constrained Adversarial Skin Lesion Image Synthesis
Kumar Abhishek, Ghassan Hamarneh
Workshop on Simulation and Synthesis in Medical Imaging (SASHIMI), International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), 2019

We propose a GAN-based synthesis approach for generating realistic skin lesion images from lesion masks, making it an appropriate augmentation strategy for skin lesion segmentation datasets.


Improved Inference via Deep Input Transfer
Saeid Asgari Taghanaki, Kumar Abhishek, Ghassan Hamarneh
International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), 2019   (Early Accept)

We propose an input image transformation technique that relies on the gradients of a trained segmentation network to transform the images for improved segmentation performance.


CloudMaskGAN: A Content-Aware Unpaired Image-to-Image Translation Algorithm for Remote Sensing Imagery
Sorour Mohajerani, Reza Asad, Kumar Abhishek, Neha Sharma, Alysha van Duynhoven, Parvaneh Saeedi
IEEE International Conference on Image Processing (ICIP), 2019

We propose an unpaired image-to-image translation algorithm for generating synthetic remote sensing images with different land cover types while preserving the locations and the intensity values of the cloud pixels.


A Kernelized Manifold Mapping to Diminish the Effect of Adversarial Perturbations
Saeid Asgari Taghanaki, Kumar Abhishek, Shekoofeh Azizi, Ghassan Hamarneh
IEEE International Conference on Computer Vision and Pattern Recognition (CVPR), 2019

We propose a non-linear radial basis convolutional feature mapping based adversarial defense that is resilient to gradient and non-gradient based attacks while also not affecting the performance of clean data.


Summarization and Visualization of Large Volumes of Broadcast Video Data
Undergraduate Thesis

[Abstract] [bibtex]

A Minutiae Count Based Method for Fake Fingerprint Detection
Kumar Abhishek, Ashok Yogi
Procedia Computer Science, Volume 58, 2015
[Abstract] [bibtex]

Non-Invasive Measurement of Heart Rate and Hemoglobin Concentration through Fingertip
Kumar Abhishek, Amodh Kant Saxena, Ramesh Kumar Sonkar
IEEE International Conference on Signal Processing, Informatics, Communication and Energy Systems (SPICES), 2015   (Oral Presentation)
[Abstract] [Presentation] [bibtex]

An Enhanced Algorithm for the Quantification of Human Chorionic Gonadotropin (hCG) Level in Commercially Available Home Pregnancy Test Kits
Kumar Abhishek, Mrinal Haloi, Sumohana S. Channappayya, Siva Rama Krishna Vanjari, Dhananjaya Dendukuri, Swathy Sridharan, Tripurari Choudhary, Paridhi Bhandari
IEEE Twentieth National Conference on Communications (NCC), 2014   (Oral Presentation)
[Abstract] [Presentation] [bibtex]

Template courtesy: Jon Barron