Profile

Cover photo
Tomasz Malisiewicz
Works at vision.ai
Attended Carnegie Mellon University
2,802 followers|8,165,756 views
AboutPostsPhotosYouTube

Stream

Tomasz Malisiewicz

Shared publicly  - 
 
Some important take-home lessons from NIPS 2016. Quite relevant to those applying machine learning in industry and startups, especially those trying to capitalize on new advancements in deep learning.
You might go to a cutting-edge machine learning research conference like NIPS hoping to find some mathematical insight that will help you take your deep learning system's performance to the next level. Unfortunately, as Andre...
1
Add a comment...

Tomasz Malisiewicz

Shared publicly  - 
 
Dropout allows us to convert neural network outputs into probabilities, with no cost to performance, and minimal computational overhead. Bayesian Deep Learning, here we come?
In Quantum Mechanics, Heisenberg's Uncertainty Principle states that there is a fundamental limit to how well one can measure a particle's position and momentum. In the context of machine learning systems, a similar principle...
31
11
Ankur Handa's profile photoYarin G's profile photo
2 comments
Yarin G
 
Thanks! 
Add a comment...
 
I wrote a lengthy blog post about what's new in the world of SLAM algorithms, together with summaries of the December 2015 Future of Real-Time SLAM workshop. I hope you guys find it educational.

http://www.computervisionblog.com/2016/01/why-slam-matters-future-of-real-time.html

#robotics #computervision #iccv #research
Last month's International Conference of Computer Vision (ICCV) was full of Deep Learning techniques, but before we declare an all-out ConvNet victory, let's see how the other "non-learning" geometric side of computer vision ...
27
7
Helio Perroni Filho's profile photo
 
It was very informative, thank you. I was especially intrigued by the article on Placeless Place Recognition, it bears some resemblance to my own research on visual self-localization, albeit coming from a completely different direction.
Add a comment...

Tomasz Malisiewicz

Shared publicly  - 
39
32
Sudeep Pillai's profile photoTomasz Malisiewicz's profile photoAvram Golbert's profile photoAndrew Davison's profile photo
6 comments
 
Yes, those papers are very interesting... Also FlowNet. There will soon be much more of this type of geometrical estimation via deep learning. Interesting question is whether we will need real SLAM data to train them or if you can do ist of the training with synthetic data as in FlowNet.
Add a comment...

Tomasz Malisiewicz

Shared publicly  - 
 
The Deep Learning Gold Rush of 2015
In the last few decades, we have witnessed major technological innovations such as personal computers and the internet finally reach the mainstream. And with mobile devices and social networks on the rise, we're now more connected than ever. So what's next?...
In the last few decades, we have witnessed major technological innovations such as personal computers and the internet finally reach the mainstream. And with mobile devices and social networks on the rise, we're now more conn...
15
5
Jason Fulton's profile photo
 
Nvidia Jetson TX1 developer module for those on more limited budgets
Add a comment...

Tomasz Malisiewicz

Shared publicly  - 
 
watch me train a gesture or two
This Hangout On Air is hosted by Tomasz Malisiewicz. The live video broadcast will begin soon.
Q&A
Preview
Live
training some characters in red rock coffee, mountain view CA
Thu, October 15, 2015, 11:17 PM
Hangouts On Air - Broadcast for free

1
Add a comment...

Tomasz Malisiewicz

Shared publicly  - 
 
Nuts and Bolts of Building Deep Learning Applications: Ng @ NIPS2016
You might go to a cutting-edge machine learning research conference like NIPS hoping to find some mathematical insight that will help you take your deep learning system's performance to the next level. Unfortunately, as Andrew Ng reiterated to a live crowd ...
You might go to a cutting-edge machine learning research conference like NIPS hoping to find some mathematical insight that will help you take your deep learning system's performance to the next level. Unfortunately, as Andre...
2
4
Add a comment...

Tomasz Malisiewicz

Shared publicly  - 
 
Making Deep Networks Probabilistic via Test-time Dropout
In Quantum Mechanics, Heisenberg's Uncertainty Principle states that there is a fundamental limit to how well one can measure a particle's position and momentum . In the context of machine learning systems, a similar principle has emerged, but relating inte...
In Quantum Mechanics, Heisenberg's Uncertainty Principle states that there is a fundamental limit to how well one can measure a particle's position and momentum. In the context of machine learning systems, a similar principle...
9
18
Add a comment...

Tomasz Malisiewicz

Shared publicly  - 
 
Deep Learning Trends @ ICLR 2016
Started by the youngest members of the Deep Learning Mafia [1], namely  Yann LeCun and Yoshua Bengio , the ICLR conference is quickly becoming a strong contender for the single most important venue in the Deep Learning space . More intimate than NIPS and le...
Started by the youngest members of the Deep Learning Mafia [1], namely Yann LeCun and Yoshua Bengio, the ICLR conference is quickly becoming a strong contender for the single most important venue in the Deep Learning space. M...
36
26
Ankur Handa's profile photoTomasz Malisiewicz's profile photo
2 comments
 
I agree, ICLR 2016 was quite memorable.
Add a comment...

Tomasz Malisiewicz

Shared publicly  - 
 
Anybody have the PDF slides from the BMVC Visual SLAM Tutorial that was given back in 2007?

https://www.cs.bris.ac.uk/Research/Vision/Realtime/bmvctutorial/

I'd like to look at the presentation slides for archival purposes (to basically see how much SLAM changed over the past 8 years), and the links to 2/3 presentations seem to be down. I can find PDFs of part II and part II when I do some serious Googling -- does anybody have part I? +Andrew Davison
Visual SLAM tutorial BMVC 2007
5
1
Andrew Davison's profile photoTomasz Malisiewicz's profile photoWalterio Mayol's profile photo
5 comments
 
Hi +Tomasz Malisiewicz  great effort to compile how things are changing.
Please try this and BTW thanks for pointing out the previous links are broken. If this one does not work please let me know:

https://drive.google.com/a/bristol.ac.uk/folderview?id=0B-LKlzV8Oy8XN3M0bmQ5aXZONFk&usp=sharing#grid

Walterio
Add a comment...

Tomasz Malisiewicz

Shared publicly  - 
 
ICCV 2015: Twenty one hottest research papers
"Geometry vs Recognition" becomes ConvNet-for-X Computer Vision used to be cleanly separated into two schools: geometry and recognition . Geometric methods like structure from motion and optical flow usually focus on measuring objective real-world quantitie...
40
16
Add a comment...

Tomasz Malisiewicz

Shared publicly  - 
 
 
As Deep Neural Network (DNN) applications grow in importance in various areas including internet search engines and medical imaging, Intel teams are working on software solutions to accelerate these workloads that will become available in future versions of Intel® Math Kernel Library (Intel® MKL) and Intel® Data Analytics Acceleration Library (Intel® DAAL). This technical preview demonstrates performance that is possible to achieve on Intel platf...
View original post
3
1
Leonid Boytsov's profile photo
 
Uhhh, wow, I guess 4x is about a speed-up that you gain from using a GPU. Here is the source of many GPU vastly outperforms CPU comparisons. You take unoptimized CPU code, carefully rewrite it for GPU. Plus, of course, peak GPUs performance is much higher (but GPUs are hard to load fully).

I wonder what will happen when the current Intel co-processors become main CPUs. I have heard this will happen quite soon. Instead of 8-16 CPU cores, we will have about 100.
Add a comment...
Story
Tagline
Computer Vision Entrepreneur
Introduction
I'm an Entrepreneur, Scientist, and the Co-Founder of vision.ai. Previously, I was a Postdoctoral Scholar at MIT's Computer Science and Artificial Intelligence Laboratory, obtained a PhD in Robotics from Carnegie Mellon University, and studied Physics/CS as an undergrad.
Bragging rights
NSF Graduate Research Fellowship (2006-2009) during my PhD at CMU
Education
  • Carnegie Mellon University
    Robotics PhD, 2008 - 2011
  • Carnegie Mellon University
    Robotics MS, 2005 - 2008
  • Rensselaer Polytechnic Institute
    Physics / Computer Science BS, 2001 - 2005
  • Patchogue-Medford High School
    High School
Work
Occupation
Computer Vision Entrepreneur and Researcher
Skills
Computer Vision, Object Detection, Image Recognition, Machine Learning, Deep Learning, Robotics, Software Engineering, Big Data, Applied Mathematics
Employment
  • vision.ai
    Cofounder, 2013 - present
    Our goal at vision.ai is to make computer vision tools and technologies accessible to the whole world. Our first product, VMX, is designed to bring cutting-edge computer vision technology to: hobbyists, researchers, artists, students, roboticists, engineers, and entrepreneurs. We are currently launching a Kickstarter campaign, so please check us out at: http://kickstarter.vmx.ai
  • Massachusetts Institute of Technology (MIT)
    Postdoc, 2011 - 2013
    I was a Postdoctoral Research Scholar at the Computer Science and Artificial Intelligence Laboratory (CSAIL) at MIT where I worked in Antonio Torralba's Lab on research in object detection, 3D scene reconstruction, and machine learning.
  • Carnegie Mellon University (CMU)
    PhD Student, 2005 - 2011
    I was a PhD student at Carnegie Mellon University's Robotics Institute where I worked with Alexei A. Efros on research in image understanding, object recognition, image segmentation, and the machine learning techniques required to solve such large-scale problems.
  • Google
    Software Engineer, 2008 - 2008
    I spent a summer working with Thomas Leung at Google Research in Mountain View, CA where I worked on an automatic segmentation algorithm for large-scale category-level image recognition. I gained experience working in a fast-paced software engineering environment and deploying computations using Google's MapReduce.
  • Google
    Software Engineer, 2009 - 2009
    I spent a summer working with Dennis Strelow at Google Research in Mountain View, CA where I worked on a discriminative groups sparse coding algorithm. I gained experience working on sparse coding problems, formulating hybrid objective functions which utilize intuitions from unsupervised learning and discriminative methods, as well as decoupling such problems for deployment in a MapReduce-heavy workflow.
  • Brookhaven National Laboratory
    Summer Intern, 2001 - 2001
    I worked on a summer project at BNL's Physics Department during which I studied numerical integration routines for modeling relativistic muons inside electromagnetic storage rings.
Basic Information
Gender
Male
Other names
Tom Malisiewicz, tombone, quantombone, t0|\/|b0|\|3