Posts by Collection
Today I’m happy to announce that I’ve started as an Assistant Professor at York University.
I am now a member of the Centre for Vision Research at York University.
A major focus of my research over the last few years as been into new methods for the determination of 3D structures of viruses and proteins. A big piece of that work has been published today in Nature Methods. This article describes the algorithmic advances which underpin cryoSPARC. This work was done in collaboration with Ali Punjani (PhD Student, University of Toronto), John Rubinstein (Senior Research Scientist, Hospital for Sick Children) and David Fleet (Professor, University of Toronto).
For a non-technical introduction to our work, you can take a look at this story out of University of Toronto, Scarborough.
Paper at the 2017 IEEE Winter Conference on Applications of Computer Vision (WACV) on predicting the malignancy of lung nodules in CT scans with 3D convolutional neural networks. Find the paper here.
Recent work on using semantic cues to localize a vehicle will be presented at the 2017 IEEE International Conference on Robotics and Automation (ICRA). This is joint work with Wei-Chui Ma, Shenlong Wang, Sanja Fidler and Raquel Urtasun. You can find the paper here and a video of some of the results here.
I also recently visited the Program in Applied and Computational Mathematics at Princeton University to deliver the IDeAS Seminar on April 27th. I’ll next be visiting and speaking at several places in Montreal and Lausanne this month. Feel free to say hello!
- May 10th - Speaking at the Canadian Microscopy and Cytometry Symposium in Montreal. See the symposium website for the presentation schedule.
- May 12th - Visiting Joaquin Ortega and Kaleem Sadiqqi at McGill University in Montreal. I’ll be giving a seminar for a computer science audience in the afternoon.
- May 15th - Speaking at the 2017 Cryo-EM Symposium at EPFLs BioEM Facility in Lausanne, Switzerland
- May 16th - Visiting the CVLab at EPFL. I’ll be giving a computer science oriented seminar in the morning.
Our new paper on dynamic textures is now available on arXiv here and the website with more results and details can be found here. Also, I’ll be attending CVPR in July. If you’re going to be there and would like to meet up, feel free to drop me an email.
I will be serving as one of the Student Volunteer Chairs for CVPR 2018.
I will be speaking on “Bayesian Methods in Cryo-EM” at the 2017 NRAMM Workshop on Advanced Topics in EM Structure Determination: Challenges and Opportunities.
I am excited to announce that I will be joining Borealis AI as a Researcher Director. This will be part time for now and become full time starting May 1st, when I take a leave of absence from my position at York University. Read more about Borealis and the work I’ll be doing there here.
I wrote a brief essay on striving for impact in research, in particular in machine learning. You can read it on TechVibes here.
I have returned full time to academia! At the moment I am working on expanding my lab and restarting my research program. Watch this space for updates in the near future!
I recently co-authored a review article (preprint version) on Normalizing Flows which will be published in IEEE Transactions of Pattern Analysis and Machine Intellgience. Normalizing Flows is a topic that I’m excited about and looking to work more on in the future. If you’ve not seen it before, please check it out.
Our paper on the “Tails of Lipschitz Triangular Flows” has been accepted for publication at ICML 2020. If you’re attending, be sure to stop by and chat.
I’ve finally launched a new website based on GitHub Pages. This should make it easier for me to maintain and add content. Keep an eye on things over the coming weeks as I work to expand and improve its contents. My old York EECS website is still availble here and my ancient UofT DCS website is still available here, although I am planning to take down my DCS page soon.
Two papers on normalizing flows have been accepted at NeurIPS 2020. Congratulations and thanks to the students, Jason and Ruizhi, and my excellent collaborators!
- Wavelet Flow: Fast Training of High Resolution Normalizing Flows with Jason Yu and Konstantinos Derpanis (arxiv preprint to come)
- Modeling Continuous Stochastic Processes with Dynamic Normalizing Flows with Ruizhi Deng, Bo Chang, Greg Mori and Andreas Lehrmann
The code and paper for our NeurIPS 2020 paper, Wavelet Flow: Fast Training of High Resolution Normalizing Flows, is now available. The code is availble through the Wavelet Flow GitHub Project. The paper can be found on arXiv. All of this, and additional image results, can be found on the Wavelet Flow Project Page. If you have questions, feel free to get in touch or check us out at our poster session at NeurIPS 2020.
This summer I presented a tutorial at ECCV 2020 on normalizing flows with Ullrich Koethe and Carsten Rother. I’ve finally uploaded my lecture which provides an introduction to normalizing flows. Feedback is welcome!
Happy to announce that my Faculty Affiliate status at the Vector Institute has been renewed!
Happy to announce that Ullrich Koethe and I will be again running our tutorial on Normalizing Flows and Invertible Neural Networks in Computer Vision at CVPR 2021. This will be an updated version of the tutorial we presented at ECCV 2020. The introductory video from that offering is available here. Feel like something should have been presented differently? Want to see more about some other topics? Let me know, feedback is welcome!
I recently wrote a tutorial on differential privacy. The first part is available now and the second part is coming soon.
Proud to announce that I’m serving as the General Chair for the first ever Ontario Workshop on Computer Vision. This is a trainee focused event aiming to provide a forum for computer vision researchers in the province of Ontario to share their early work, network and establish new collaborations. There will be several keynote sessions, poster sessions and more. Keep an eye on the website and OWCV twitter accounts for more information.
HistoGAN: Controlling Colors of GAN-Generated and Real Images via Color Histograms with Michael Brown and Mahmoud Afifi has been accepted at CVPR 2021. Congratulations and thanks to the student, Mahmoud Afifi, and his supervisor Michael Brown for the fruitful collaboration!
The second part of my tutorial on differential privacy is now available. The first part was a basic introduction to DP, while part II focuses on using differential privacy in the context of machine learning.
Happy to announce that I will be giving a seminar on Normalizing Flows in Theory and Practice as part of the CAIDA seminar series at the University of British Columbia. For more information and event registration, check out the event announcement. There are likely to be more seminars coming in the near future, keep an eye here for announcements.
Congratulations to my student, Shayan Kousha, whose paper Zero-shot Learning with Class Description Regularization was accepted to the Fine-Grained Visual Categorization Workshop at CVPR 2021. Stay tuned here for more info once the paper is released.
Happy to annouce that I will be giving an invited talk at the INNF+ Workshop at ICML this year. My collaborators and I also had two papers accepted for presentation at the workshop:
- Manifold Density Estimation via Generalized Dequantization by James A Brofos, Marcus A Brubaker and Roy R Lederman
- Agent Forecasting at Flexible Horizons using ODE Flows by Alexander Radovic, Jiawei He, Janahan Ramanan, Marcus A Brubaker and Andreas Lehrmann (paper to come)
I am excited to announce that I have joined the Toronto Samsung AI Center part time as a Visiting Professor. I will be overseeing projects in computer vision and machine learning with their excellent team of researchers. If you’re looking for an industrial position, be sure to check them out!
Several new workshop papers and preprints are now available on arXiv!
- CAMS: Color-Aware Multi-Style Transfer by Mahmoud Afifi, Abdullah Abuolaim, Mostafa Hussien, Marcus A. Brubaker and Michael S. Brown, preprint.
- Continuous Latent Process Flows by Ruizhi Deng, Marcus A. Brubaker, Greg Mori and Andreas Lehrmann at ICML Workshop on Time Series, 2021.
- Manifold Density Estimation via Generalized Dequantization by James A. Brofos, Marcus A. Brubaker and Roy R. Lederman at ICML Workshop on Invertible Neural Networks, Normalizing Flows, and Explicit Likelihood Models, 2021.
- Zero-shot Learning with Class Description Regularization by Shayan Kousha and Marcus A. Brubaker at CVPR Workshop on Fine-Grained Visual Categorization, 2021.
A new preprint is now available on arXiv!
- Neural Image Representations for Multi-Image Fusion and Layer Separation by Seonghyeon Nam, Marcus A. Brubaker and Michael S. Brown, preprint.
I’m very happy to announce that my lab has been awarded a Canada Foundation for Innovation grant as part of the John R. Evans Leaders Fund. The grant, valued at $140,000CAD, will be used to purchase hardware to continue my groups research on generative models with a focus on applications in cryo-EM, hyperspectral imagery and video analysis.
I will be giving a seminar about my work on normalizing flows on Friday at Amii. More information, including timing and registration, can be found here. Can’t make this one? There are several more seminars in the works for the next few months.
I’m happy to announce that I’ll be serving as a Senior Area Chair for AAAI 2022.
Happy to announce two newly accepted papers. Congratulations to all my co-authors!
- Continuous Latent Process Flows by Ruizhi Deng, Marcus A. Brubaker, Greg Mori and Andreas Lehrmann at NeurIPS 2021.
- Auto White-Balance Correction for Mixed-Illuminant Scenes by Mahmoud Afifi, Marcus A. Brubaker and Michael S. Brown at WACV 2022.
Happy to announce another newly accepted paper. Adaptation of the Independent Metropolis-Hastings Sampler with Normalizing Flow Proposals by James A. Brofos, Marylou Gabrie, Marcus A. Brubaker and Roy R. Lederman will appear at AISTATS 2022. Congratulations to all my co-authors!
Happy to announce three accepted papers at CVPR this year! They are currently not yet available but will be made available soon.
The following new papers are now available:
- Noise2NoiseFlow: Realistic Camera Noise Modeling without Clean Images by Ali Maleky, Shayan Kousha, Michael S Brown, Marcus A Brubaker at CVPR 2022.
- Modeling sRGB Camera Noise with Normalizing Flows by Shayan Kousha, Ali Maleky, Michael S Brown, Marcus A Brubaker at CVPR 2022.
- Learning sRGB-to-Raw De-rendering with Content-Aware Metadata by Seonghyeon Nam and Abhijith Punnappurath and Marcus A. Brubaker and Michael S. Brown at CVPR 2022.
- Residual Multiplicative Filter Networks for Multiscale Reconstruction by Shayan Shekarforoush, David B Lindell, David J Fleet, Marcus A Brubaker on arXiv.
- Efficient CDF Approximations for Normalizing Flows by Chandramouli Shama Sastry, Andreas Lehrmann, Marcus Brubaker, Alexander Radovic on arXiv.
Our paper Neural Image Representations for Multi-Image Fusion and Layer Separation by Seonghyeon Nam, Marcus A. Brubaker and Michael S. Brown, has been accepted for publication at ECCV 2022.
Our paper Efficient CDF Approximations for Normalizing Flows by Chandramouli Shama Sastry, Andreas Lehrmann, Marcus Brubaker, Alexander Radovic has been accepted for publication at Transactions on Machine Learning Research (TMLR). Congratulations and thanks to all my excellent collaborators!
Our paper Residual Multiplicative Filter Networks for Multiscale Reconstruction by Shayan Shekarforoush, David B Lindell, David J Fleet, Marcus A Brubaker on arXiv. has been accepted for publication at NeurIPS 2022. Congratulations to the student Shayan on his first paper and thanks to all the collaborators!