Are you over 18 and want to see adult content?
More Annotations
![A complete backup of narbonne-tourisme.com](https://www.archivebay.com/archive2/59c27703-2f44-4956-9e0c-79234926ea44.png)
A complete backup of narbonne-tourisme.com
Are you over 18 and want to see adult content?
![A complete backup of autisme-ambitionavenir.com](https://www.archivebay.com/archive2/c01df845-2a29-4ea5-9370-6f4060c1525b.png)
A complete backup of autisme-ambitionavenir.com
Are you over 18 and want to see adult content?
![A complete backup of codeconquest.com](https://www.archivebay.com/archive2/4f870aa3-a282-4ab8-bdd4-06b850b3cad6.png)
A complete backup of codeconquest.com
Are you over 18 and want to see adult content?
![A complete backup of remingtoncollege.edu](https://www.archivebay.com/archive2/ced12547-9b9f-43b1-9c11-e56750af10e5.png)
A complete backup of remingtoncollege.edu
Are you over 18 and want to see adult content?
![A complete backup of daisydiskapp.com](https://www.archivebay.com/archive2/eb7705e0-b57d-486e-b093-caa4eef4059c.png)
A complete backup of daisydiskapp.com
Are you over 18 and want to see adult content?
![A complete backup of reese-hitches.com](https://www.archivebay.com/archive2/5a803ee1-29a9-446c-9c8d-b42eb51d2844.png)
A complete backup of reese-hitches.com
Are you over 18 and want to see adult content?
Favourite Annotations
![A complete backup of drivingtestnsw.com](https://www.archivebay.com/archive2/bdab19fa-30eb-4af2-8a00-9b98d3f03f64.png)
A complete backup of drivingtestnsw.com
Are you over 18 and want to see adult content?
![A complete backup of fukuroecchi.tumblr.com](https://www.archivebay.com/archive2/6cb85339-bcbb-41fa-8348-4b0823201190.png)
A complete backup of fukuroecchi.tumblr.com
Are you over 18 and want to see adult content?
![A complete backup of yonuncanuncahe.blogspot.com](https://www.archivebay.com/archive2/ae4fdda3-08d5-4812-b795-47d1752ccd4a.png)
A complete backup of yonuncanuncahe.blogspot.com
Are you over 18 and want to see adult content?
![A complete backup of offbroadwaystl.com](https://www.archivebay.com/archive2/c87e5e73-e3de-4e61-bb76-90090fefa72b.png)
A complete backup of offbroadwaystl.com
Are you over 18 and want to see adult content?
![A complete backup of wyszukiwarkamp3.com.pl](https://www.archivebay.com/archive2/9348b0ae-3595-4de9-a24b-396fc48616f8.png)
A complete backup of wyszukiwarkamp3.com.pl
Are you over 18 and want to see adult content?
![A complete backup of bastl-instruments.com](https://www.archivebay.com/archive2/f20ebdaf-1e17-4ff7-b9ca-8122aebedf51.png)
A complete backup of bastl-instruments.com
Are you over 18 and want to see adult content?
![A complete backup of whiteplainschrysler.com](https://www.archivebay.com/archive2/317d8161-567d-4f74-b4e4-cd591a2977a7.png)
A complete backup of whiteplainschrysler.com
Are you over 18 and want to see adult content?
Text
BORIS BABENKO, PHD
about i'm a machine learning / computer vision engineer. i received a PhD from UCSD where I was advised by serge belongie.most of my graduate work was funded by an NSF IGERT traineeship and a 2010 google fellowship, and revolved around weakly supervised learning and its applications to object detection, recognition and tracking; one of my algorithms is included in OpenCV 3.0.PUBLICATIONS
Boris Babenko, Siva Balasubramanian, Katy E. Blumer, Greg S. Corrado, Lily Peng, Dale R. Webster, Naama Hammel, Avinash V. Varadarajan arXiv, 2019 @misc {1904.05478, Author = {Boris Babenko and Siva Balasubramanian and Katy E. Blumer and Greg S. Corrado and Lily Peng and Dale R. Webster and Naama Hammel and Avinash V PRECISION AND RECALLBORIS BABENKO, PHD
"Robust Object Tracking with Online Multiple Instance Learning" Boris Babenko, Ming-Hsuan Yang, Serge Belongie IEEE TPAMI, August 2011 @inproceedings {babenko11, title = {Robust Object Tracking with Online Multiple Instance Learning}, journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)}, year = {2011}, author = {Boris Babenko and Ming-Hsuan Yang Serge Belongie}} WEIGHT DECAY VS L2 REGULARIZATION the key difference is the pesky factor of 2! so, if you had your weight decay set to 0.0005 as in the AlexNet paper and you move to a deep learning framework that implements L2 regularization instead, you should set that \ (\lambda\) hyperparameter to 0.0005/2.0 to get the same behavior. this is what ended up causing the difference whenmoving
CONVOLUTIONAL LEARNINGS: THINGS I LEARNED BY IMPLEMENTING deep learning has taken the machine learning world by storm, achieving state of the art results in a number of applications. whether neural nets are here to stay or will be replaced be the next hot thing two years from now, one thing is certain: they are now a critical component in any machine learning expert’s toolbox (along with svm, random forests, etc). TEMPLATE MATCHING BASICS the final step is to run template matching (normalized cross correlation), and find the peaks in the response (aka non-maximal suppression), getting you the rough locations of the beads: results are not perfect – some beads on the sides of the image didn’t get detected, and the locations are probably not perfectly in the centerof each bead
HARRY POTTER AND THE CURSE OF DIMENSIONALITY john von nuemann was once quote to say that “in mathematics you don’t understand things, you just get used to them” (this quote hung on my monitor throughout grad school). ROBUST OBJECT TRACKING WITH ONLINE MULTIPLE INSTANCE LEARNING Robust Object Tracking with Online Multiple Instance Learning Boris Babenko, Student Member, IEEE, Ming-Hsuan Yang, Senior Member, IEEE and Serge Belongie, Member, IEEE AUTOMATED ANALYSIS OF PIN-4 STAINED PROSTATE NEEDLE BIOPSIES Automated Analysis of PIN-4 Stained Prostate Needle Biopsies 91 needed for creatingimagingobjects and processingthem. Image processingfunc-tions that are available in the off-the shelf commercial and open source systemsBORIS BABENKO, PHD
about i'm a machine learning / computer vision engineer. i received a PhD from UCSD where I was advised by serge belongie.most of my graduate work was funded by an NSF IGERT traineeship and a 2010 google fellowship, and revolved around weakly supervised learning and its applications to object detection, recognition and tracking; one of my algorithms is included in OpenCV 3.0.PUBLICATIONS
Boris Babenko, Siva Balasubramanian, Katy E. Blumer, Greg S. Corrado, Lily Peng, Dale R. Webster, Naama Hammel, Avinash V. Varadarajan arXiv, 2019 @misc {1904.05478, Author = {Boris Babenko and Siva Balasubramanian and Katy E. Blumer and Greg S. Corrado and Lily Peng and Dale R. Webster and Naama Hammel and Avinash V PRECISION AND RECALLBORIS BABENKO, PHD
"Robust Object Tracking with Online Multiple Instance Learning" Boris Babenko, Ming-Hsuan Yang, Serge Belongie IEEE TPAMI, August 2011 @inproceedings {babenko11, title = {Robust Object Tracking with Online Multiple Instance Learning}, journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)}, year = {2011}, author = {Boris Babenko and Ming-Hsuan Yang Serge Belongie}} WEIGHT DECAY VS L2 REGULARIZATION the key difference is the pesky factor of 2! so, if you had your weight decay set to 0.0005 as in the AlexNet paper and you move to a deep learning framework that implements L2 regularization instead, you should set that \ (\lambda\) hyperparameter to 0.0005/2.0 to get the same behavior. this is what ended up causing the difference whenmoving
CONVOLUTIONAL LEARNINGS: THINGS I LEARNED BY IMPLEMENTING deep learning has taken the machine learning world by storm, achieving state of the art results in a number of applications. whether neural nets are here to stay or will be replaced be the next hot thing two years from now, one thing is certain: they are now a critical component in any machine learning expert’s toolbox (along with svm, random forests, etc). TEMPLATE MATCHING BASICS the final step is to run template matching (normalized cross correlation), and find the peaks in the response (aka non-maximal suppression), getting you the rough locations of the beads: results are not perfect – some beads on the sides of the image didn’t get detected, and the locations are probably not perfectly in the centerof each bead
HARRY POTTER AND THE CURSE OF DIMENSIONALITY john von nuemann was once quote to say that “in mathematics you don’t understand things, you just get used to them” (this quote hung on my monitor throughout grad school). ROBUST OBJECT TRACKING WITH ONLINE MULTIPLE INSTANCE LEARNING Robust Object Tracking with Online Multiple Instance Learning Boris Babenko, Student Member, IEEE, Ming-Hsuan Yang, Senior Member, IEEE and Serge Belongie, Member, IEEE AUTOMATED ANALYSIS OF PIN-4 STAINED PROSTATE NEEDLE BIOPSIES Automated Analysis of PIN-4 Stained Prostate Needle Biopsies 91 needed for creatingimagingobjects and processingthem. Image processingfunc-tions that are available in the off-the shelf commercial and open source systems LEARNING LOW-LEVEL VISION FEAUTRES IN ~10 LINES OF CODE adam coates and colleagues have a string of very interesting papers where they propose using k-means to train convolutional networks. training the first layer boils down to about 10 lines of python code (thanks to sklearn’s implementation of k-means).BBABENKO.GITHUB.IO
As you ski down the slopes, you must avoid these obstacles: TEMPLATE MATCHING BASICS the final step is to run template matching (normalized cross correlation), and find the peaks in the response (aka non-maximal suppression), getting you the rough locations of the beads: results are not perfect – some beads on the sides of the image didn’t get detected, and the locations are probably not perfectly in the centerof each bead
CONVOLUTIONAL LEARNINGS: THINGS I LEARNED BY IMPLEMENTING deep learning has taken the machine learning world by storm, achieving state of the art results in a number of applications. whether neural nets are here to stay or will be replaced be the next hot thing two years from now, one thing is certain: they are now a critical component in any machine learning expert’s toolbox (along with svm, random forests, etc). MISC - BBABENKO.GITHUB.IO gifs. blue through red. periodically, i write and record music in my living room.. carven von bearensquash. my alter ego and the author ofthis seminal paper.
WHEN HOGS PY
as you can see, the dimensionality checks out. also, looks like opencv is about 30x faster. the skimage implementation is written in pure python (i.e. not cython), so a 30x difference is about what one would expected between a python and c++ implementations.. now, are the outputs of these two implementations the same? MULTIPROCESSING AND SEEDED RNGS seems to work ok, but it’s a bit slow. if your computer has a multi-core cpu, it’d be nice to leverage all the cores. python threads won’t help you with this, though, because of the GIL (global interpreter lock). luckily, there’s a handy python module called multiprocessing – it effectively spawns multiple python interpreters for you, e.g. one for each core, so that you can HARRY POTTER AND THE CURSE OF DIMENSIONALITY john von nuemann was once quote to say that “in mathematics you don’t understand things, you just get used to them” (this quote hung on my monitor throughout grad school). MOVING MY WEBSITE AND BLOG TO GITHUB i finally bit the bullet and moved my website and blog to github pages.all in all, it was fairly straight forward, and the jekyll backend offers a lot of nice features for technical blog writing: DEAR ZAPPOS: HERE’S A WAY YOU CAN SIGNIFICANTLY IMPROVE i bought two pairs of shoes on zappos the other day. they were both from the same manufacturer (vans), and the same size. one pair fit, the other one didn’t.boris babenko, phd
publications misc
ABOUT
i'm a machine learning / computer vision engineer. i received a PhD from UCSD where I was advised by serge belongie. most of my
graduate work was funded by an NSF IGERT traineeship and a 2010 googlefellowship
,
and revolved around weakly supervised learning and its applications to object detection, recognition and tracking; one of my algorithms is included in OpenCV 3.0.
after finishing my PhD, I started a company called anchovi labs which was acquired by dropbox;
i worked at dropbox for 2.5 years. currently i work at orbital insight where i develop computer vision algorithms to analyze satellite imagery. oh, and if you haven't noticed already, i have no respect for capital letters. in my spare time i play guitar, read/watch sci-fi, and occasionallyblog/tweet.
BLOG POSTS
* 2018-04-27 - weight decay vs L2 regularization * 2018-04-23 - practical tips (life, machine learning, andeverything)
* 2017-10-01 - precision and recall * 2017-08-25 - moving my website and blog to github * 2016-03-27 - multiprocessing and seeded RNGs * 2015-01-03 - deep net highlights from 2014 * 2014-11-29 - template matching basics * 2014-05-24 - learning low-level vision feautres in ~10 lines ofcode
* 2014-04-20 - convolutional learnings: things i learned by implementing convolutional neural nets * 2013-08-12 - dear zappos: here’s a way you can significantly improve user experience * 2013-07-28 - when hogs py * 2013-05-13 - harry potter and the curse of dimensionalityGitHub icon
Details
Copyright © 2024 ArchiveBay.com. All rights reserved. Terms of Use | Privacy Policy | DMCA | 2021 | Feedback | Advertising | RSS 2.0