Md Amirul Islam

I am a first year PhD student at the University of Manitoba working with Dr. Neil Bruce at Computer Vision Lab. My research interests are in the fields of computer vision and deep learning, with a focus on semantic analysis and understanding of images. I aim to find efficient solutions for problems that are grounded in applications such as Autonomous Driving.

I received my MSc in Computer Science from the University of Manitoba in 2017, where I was advised by Dr. Yang Wang and Dr. Neil Bruce in the Computer Vision Lab. Previously, I did my bachelors in Computer Science and Engineering at North South University.

Research Interests: Deep Learning, Computer Vision, Visual Recognition, Dense Image Labeling.

Email  /  CV  /  Google Scholar  /  LinkedIn  /  GitHub

Research
Gated Feedback Refinement Network for Dense Image Labeling
Md Amirul Islam, Mrigank Rochan, Neil D. B. Bruce, Yang Wang
Computer Vision and Pattern Recognition (CVPR), 2017
[project page]   [pdf]   [code]   [poster]  

A high degree of success may be achieved across a variety of dense image labeling tasks, using a relatively simple model structure in applying a canonical gating mechanism.

Dense Image Labeling Using Deep Learning
Md Amirul Islam    (Supervisor: Dr. Yang Wang and Dr. Neil Bruce)
Masters Thesis, University of Manitoba, 2017
[Thesis]   [slides]  

Encoder-Decoder network, Gating Mechanishm, Deep Supervision, Coarse-to-Fine Refinement

Salient Object Detection using a Context-Aware Refinement Network
Md Amirul Islam, Mahmoud Kalash, Mrigank Rochan, Neil D. B. Bruce, Yang Wang
British Machine Vision Conference (BMVC), 2017
[project page]   [pdf]   [poster]  

We propose an end-to-end encoder-decoder network that employs recurrent refinement to generate a saliency map in a coarse-to-fine fashion by incorporating finer details in the detection framework.

Label Refinement Network for Coarse-to-Fine Semantic Segmentation
Md Amirul Islam, Shujon Naha, Mrigank Rochan, Neil D. B. Bruce, Yang Wang
arXiv Preprint, 2017
[arxiv]  

We propose a novel network architecture called the label refinement network that predicts segmentation labels in a coarse-to-fine fashion at several resolutions.

Dense Image Labeling Using Deep Convolutional Neural Networks
Md Amirul Islam, Neil D. B. Bruce, Yang Wang
Canadian Conference on Computer and Robot Vision (CRV), 2016 (Oral)
[pdf]   [slides]  

We propose a dense image labeling approach based on DCNNs coupled with a support vector classifier.

Feature Fusion for Robust Object Tracking
Md Amirul Islam, M. Rasheduzzaman, M. M. Lutfe Elahi, Bruce Poon, M. A. Amin, Hong Yan
Int. Conference on Wavelet Analysis and Pattern Recognition (ICWAPR), 2015 (Oral)
[pdf]  

We propose a tracking framework that explicitly decomposes the long term tracking task into tracking, detection and fusion.

Workload Prediction on Google Cluster Trace
M. Rasheduzzaman, Md Amirul Islam, Rashedur M. Rahman
Journal of Grid and High Performance Computing (IJGHPC) , 6(3), 2014
[pdf]  

We design and compare different forecasting models (ANFIS, NARX, ARIMA, SVR) to predict future workload on google cluster trace.

Task Shape Classification and Workload Characterization of Google Cluster Trace
M. Rasheduzzaman, Md Amirul Islam, Tasvirul Islam, Tahmid Hossain, Rashedur M. Rahman
IEEE Int. Advance Computing Conference (IACC), 2014 (Oral)
[pdf]  

We present a simple technique for constructing workload characteristics and also provide production insights for understanding workload performance in cluster machine


This guy has an awesome website