Devinder Kumar

Devinder Kumar devinder.kumar"AT"uwaterloo.ca

PhD Candidate, University of Waterloo (Expected Graduation Dec 2019)
Student Member, Vector Institute
Lead AI scientist @NEXTAI (Next Canada)
Intel AI Ambassador
Principal Investigator (Not-for-profit), Compute Canada

My research centers around Deep Learning and its application in Computer Vision. Specifically, the research problem I am currently focusing on is: How to make current deep models interpretable (Explainable-AI).

I am a senior PhD Candidate (2017-expected graduation Dec 2019) at VIP lab - University of Waterloo and Machine Learning Research Group - University of Guelph where I am supervised by Dr. Alexander Wong (UWaterloo/ Canada research chair, Medical Imaging) and Dr. Graham Taylor (UGuelph/ Vector Inst./ Canada research chair, ML Systems). Before this, I completed my MASc. in WAVE lab & VIP lab at University of Waterloo, where I worked with Dr. Steven Waslander and Prof. David Clausi.

I am currently an ambassador with Intel AI and also serve as the Lead AI Scientist for NextAI, a world class AI accelerator program of NEXT Canada based in Toronto, Canada.

Previous to joining UWaterloo, I was a research engineer in LIP-6 - UPMC-Sorbonne University, Paris. There I had great time working with Prof. Matthieu Cord and Asst. Prof Nicolas Thome

I also frequently collaborate with: Prof. Farzad Khalvati, University of Toronto, Dr. Angelo Zilleti, FHI-Max Planck Berlin and Dr. Amarjot Singh , University of Cambridge (now at Stanford).

For a full history, here's my Curriculum-Vitæ, Google Scholar and LinkedIn.

News

Dec 2019: Paper accepted at IEEE Winter Conference on Applications of Computer Vision (WACV), 2020
Aug 2019: I gave an invited talk at SPORTLOGIQ, Montreal, Quebec on "Tips and Tricks for Traninig Neural Networks Effectively" on 26th Aug
Feb 2019: I will be giving keynote speech at the annual symposium of the Project Management Insitute (PMI) south western ontario chapter on April 25th.
Feb 2019: Gave invited talk on state of explainable AI at Element-AI and Bank of Montreal (BMO) in Toronto, Canada
Dec 2018: Awarded President Graduate Scholarship (PGS) at University of Waterloo
Dec 2018: Awarded International Ontario Graduate Scholarship (OGS) (1/5 awards at UWaterloo, 1/1 in Engg Dept.)
Aug 2018: I am chairing the Poster Session at Toronto Machine Learning Summit (TMLS), 2018. Submission of poster ideas: link
May 2018: Paper accepted at Nature Communications (IF:13.092) on interpretable crytsal lattice symmetry classification.
Dec 2017: Awarded President Graduate Scholarship (PGS) at University of Waterloo
Dec 2017: Awarded International Ontario Graduate Scholarship (OGS) (1/5 awards at UWaterloo)
Dec 2017: Our paper won Best paper Award at 31st Neural Information Processing Systems (NIPS) Transparent and interpretable Machine Learning in Safety Critical Environments Workshop, 2017
Oct 2017: Invited talk at Industrial Research Assistance Program (IRAP)-NRC meeting, ”Future of commercialization of AI research and how SMEs in Canada could exploit the opportunities in global context”, Toronto, Canada, Nov 1st
Oct 2017: I will be speaking at the Toronto Machine Learning Summit (TOML) 2017 about my X-AI research
Sept 2017: I will be speaking on the panel "Math education for an AI world" at Fields Insitute on Sept 30th
Aug 2017: I am now the managing editor for Journal of Computational Vision and Imaging Systems (JCVIS)
Jun 2017: I am giving an invited Keynote Speech at AI-Toronto, 2017 Conference on 21st June, 2017
May 2017: "CLEAR visualization" paper accepted at CVPR Workshop (CVPR-W) on Explainable Computer Vision
Apr 2017: Accepted at DL & RL Summer School, 2017 at MILA, UMontreal with CIFAR scholarship.
Mar 2017: Paper accepted at 14th International Conference on Image Analysis and Recognition (ICIAR, 2017)
Feb 2017: I am joining this year's NextAI program as an AI scientist in residence.
Dec 2016: I will be giving an invited talk at the theory department at Fritz Haber Institute of the Max Planck Society on 08 Dec, 2016
Nov 2016: Paper accepted at 30th Neural Information Processing Systems Machine learning for healthcare Workshop (NIPS MLH, 2016)
Jun 2016: I will be joining Philips Research (HQ) as a deep learning intern at Eindhoven, NL where I will be working with the data science and medical oncology solutions departments.
May 2016: I gave an invited talk at Gustave Roussy cancer research institute, Paris, France
Nov 2015: Paper accepted at 29th Neural Information Processing Systems (NIPS) Machine Learning for Healthcare Workshop ,2015

Publications selected publications only, for full list please refer my CV

2017

Explaining the Unexplained: A CLass-Enhanced Attentive Response (CLEAR) Approach to Understanding Deep Neural Networks

D.Kumar, A. Wong, and G. W. Taylor

Computer Vision and Pattern Recognition (CVPR) Workshop Oral, 2017

2016

Understanding Anatomy Classification Using Visualization

D.Kumar, and V.Menkovski

30th NIPS Machine Learning for Health (NIPS-MLH) Workshop, 2016

2015

Discovery Radiomics via StochasticNet Sequencers for Cancer Detection

M.J.Shafiee, A.G. Chung, D. Kumar,, F. Khalvati, M.A. Haider and A. Wong

29th NIPS Machine Learning for Healthcare (NIPS-MLH) Workshop, 2015

2015

| |

Recipe recognition with large multimodal food dataset

X.Wang, D.Kumar, N.Thome, M.Cord, and F.Precioso

7th Workshop on Multimediafor Cooking and Eating Activities (CEA) in conjuction with ICME, July 2015.

2015

Taming the North: Multi-Camera Parallel Tracking and Mapping in Snow-Laden Environments

A.Das, D.Kumar, A.E.Bably and S. Waslander

In proceedings Field and Service Robotics (FSR), June, 2015.

2015

| |

Lung Nodule Classification Using Deep Features in CT Images (Oral)

D. Kumar, A. Wong and D. Clausi

12th Conference on Computer and Robot Vision (CRV), Halifax, NS, June, 2015.

2013

Best Paper Award

Constructive Learning for Human-Robot Interaction

A.Singh, S.Karanam, and D.Kumar

IEEE Potentials, Vol. 32, Issue: 4, pp(s): 13-19, Aug, 2013

Software

VISIIR web demo Food recognition system based on the work that I did at LIP-6, Paris.

More projects can be found on my GitHub profile.

Data

UPMC-101 Dataset of 101 food categories (~1000 images per category) that I created from Google image search results.

Media Coverage

VICE's Motherboad: "This Researcher Wants to Open the 'Black Box' of Financial AI", Oct 2017
Canadian Underwriter: "New software could make it easier to “adopt and trust” AI systems that set insurance premiums", Oct 2017
Technology.org: "Scientists developed software to make artificial intelligence systems more trustworthy", Oct 2017
Exchange Magazine: "Waterloo research paves the way for use of complex AI in the financial sector", Oct 2017
Waterloo Media Release: "Waterloo research paves the way for use of complex AI in the financial sector", Oct 2017
Investment Executive Magazine: "Building trust in AI", Oct 2017
Engineer The Future: "Reading the minds of deep learning AI systems", Sept 2017
Waterloo Stories: "Reading the minds of deep learning AI systems", June 2017