[Colloquium] REMINDER: 6/12 TTIC Colloquium: Erik Learned-Miller, University of Massachusetts

[Mary Marre via Colloquium]

When:     Monday, June 12th at 11:00 a.m.

Where:    TTIC, 6045 S. Kenwood Avenue, 5th Floor, Room 526

Who:       Erik Learned-Miller, University of Massachusetts


Title:       Learning by Moving and by Clustering Movie Characters

Abstract:
I will present two mini-talks.

Part 1: We present an end-to-end system for detecting and clustering faces
by identity in full-length  movies. Unlike works that start with a
predefined set of detected faces, we consider the end-to-end problem of
detection and clustering together. We make three separate contributions.
First, we combine a state-of-the-art face detector with a generic tracker
to extract high quality face tracklets. We then introduce a novel
clustering method, motivated by classic results in graph theory. It is
based on the observation that  large clusters can be fully connected by
joining just a small fraction of their point pairs, while just a single
connection between two different people can lead to poor clustering
results. This suggests clustering using a verification system with {\em
very few false positives} but perhaps moderate recall. We introduce such a
verification procedure with good recall in the low false-positive regime,
based on features from the analysis of differences (FAD).  Finally, we
define a novel end-to-end detection and clustering evaluation metric
allowing us to assess the accuracy of the entire end-to-end system.  We
present state-of-the-art results on multiple video data sets and also on
standard face databases.

Part 2:  I will discuss our work in motion segmentation, based on the
geometry of perspective projection and an EM-style approach to modeling the
motion field. Our model takes as input *only the optical flow* and segments
the image into moving parts based on the optical flow. Because it does not
rely on the difference in appearance between the background and various
moving objects, it is able to handle extremely difficult cases of motion
segmentation, such as those involving camouflaged animals. I will also talk
recent efforts, collaborating with Greg Shakhnavorich, Gustav Larsson, and
Michael Maire, to use motion segmentation in an automatic supervision
scheme, where a network is trained to predict depth from a single image,
using our motion segmentation system (and subsequent calculation of depth
fields) as “automatic supervision”. We show the benefit of such an approach
for pre-training a network to solve semantic segmentation problems.


Host: Greg Shakhnarovich <greg at ttic.edu>



For more information on the colloquium series or to subscribe to the
mailing list, please see http://www.ttic.edu/colloquium.php


Mary C. Marre
Administrative Assistant
*Toyota Technological Institute*
*6045 S. Kenwood Avenue*
*Room 504*
*Chicago, IL  60637*
*p:(773) 834-1757*
*f: (773) 357-6970*
*mmarre at ttic.edu <mmarre at ttic.edu>*

On Sun, Jun 11, 2017 at 10:18 PM, Mary Marre <mmarre at ttic.edu> wrote:

> When:     Monday, June 12th at 11:00 a.m.
>
> Where:    TTIC, 6045 S. Kenwood Avenue, 5th Floor, Room 526
>
> Who:       Erik Learned-Miller, University of Massachusetts
>
>
> Title:       Learning by Moving and by Clustering Movie Characters
>
> Abstract:
> I will present two mini-talks.
>
> Part 1: We present an end-to-end system for detecting and clustering faces
> by identity in full-length  movies. Unlike works that start with a
> predefined set of detected faces, we consider the end-to-end problem of
> detection and clustering together. We make three separate contributions.
> First, we combine a state-of-the-art face detector with a generic tracker
> to extract high quality face tracklets. We then introduce a novel
> clustering method, motivated by classic results in graph theory. It is
> based on the observation that  large clusters can be fully connected by
> joining just a small fraction of their point pairs, while just a single
> connection between two different people can lead to poor clustering
> results. This suggests clustering using a verification system with {\em
> very few false positives} but perhaps moderate recall. We introduce such a
> verification procedure with good recall in the low false-positive regime,
> based on features from the analysis of differences (FAD).  Finally, we
> define a novel end-to-end detection and clustering evaluation metric
> allowing us to assess the accuracy of the entire end-to-end system.  We
> present state-of-the-art results on multiple video data sets and also on
> standard face databases.
>
> Part 2:  I will discuss our work in motion segmentation, based on the
> geometry of perspective projection and an EM-style approach to modeling the
> motion field. Our model takes as input *only the optical flow* and segments
> the image into moving parts based on the optical flow. Because it does not
> rely on the difference in appearance between the background and various
> moving objects, it is able to handle extremely difficult cases of motion
> segmentation, such as those involving camouflaged animals. I will also talk
> recent efforts, collaborating with Greg Shakhnavorich, Gustav Larsson, and
> Michael Maire, to use motion segmentation in an automatic supervision
> scheme, where a network is trained to predict depth from a single image,
> using our motion segmentation system (and subsequent calculation of depth
> fields) as “automatic supervision”. We show the benefit of such an approach
> for pre-training a network to solve semantic segmentation problems.
>
>
> Host: Greg Shakhnarovich <greg at ttic.edu>
>
>
>
> For more information on the colloquium series or to subscribe to the
> mailing list, please see http://www.ttic.edu/colloquium.php
>
>
>
> Mary C. Marre
> Administrative Assistant
> *Toyota Technological Institute*
> *6045 S. Kenwood Avenue*
> *Room 504*
> *Chicago, IL  60637*
> *p:(773) 834-1757 <(773)%20834-1757>*
> *f: (773) 357-6970 <(773)%20357-6970>*
> *mmarre at ttic.edu <mmarre at ttic.edu>*
>
> On Mon, Jun 5, 2017 at 7:46 PM, Mary Marre <mmarre at ttic.edu> wrote:
>
>> When:     Monday, June 12th at 11:00 a.m.
>>
>> Where:    TTIC, 6045 S. Kenwood Avenue, 5th Floor, Room 526
>>
>> Who:       Erik Learned-Miller, University of Massachusetts
>>
>>
>> Title:       Learning by Moving and by Clustering Movie Characters
>>
>> Abstract:
>> I will present two mini-talks.
>>
>> Part 1: We present an end-to-end system for detecting and clustering
>> faces by identity in full-length  movies. Unlike works that start with a
>> predefined set of detected faces, we consider the end-to-end problem of
>> detection and clustering together. We make three separate contributions.
>> First, we combine a state-of-the-art face detector with a generic tracker
>> to extract high quality face tracklets. We then introduce a novel
>> clustering method, motivated by classic results in graph theory. It is
>> based on the observation that  large clusters can be fully connected by
>> joining just a small fraction of their point pairs, while just a single
>> connection between two different people can lead to poor clustering
>> results. This suggests clustering using a verification system with {\em
>> very few false positives} but perhaps moderate recall. We introduce such a
>> verification procedure with good recall in the low false-positive regime,
>> based on features from the analysis of differences (FAD).  Finally, we
>> define a novel end-to-end detection and clustering evaluation metric
>> allowing us to assess the accuracy of the entire end-to-end system.  We
>> present state-of-the-art results on multiple video data sets and also on
>> standard face databases.
>>
>> Part 2:  I will discuss our work in motion segmentation, based on the
>> geometry of perspective projection and an EM-style approach to modeling the
>> motion field. Our model takes as input *only the optical flow* and segments
>> the image into moving parts based on the optical flow. Because it does not
>> rely on the difference in appearance between the background and various
>> moving objects, it is able to handle extremely difficult cases of motion
>> segmentation, such as those involving camouflaged animals. I will also talk
>> recent efforts, collaborating with Greg Shakhnavorich, Gustav Larsson, and
>> Michael Maire, to use motion segmentation in an automatic supervision
>> scheme, where a network is trained to predict depth from a single image,
>> using our motion segmentation system (and subsequent calculation of depth
>> fields) as “automatic supervision”. We show the benefit of such an approach
>> for pre-training a network to solve semantic segmentation problems.
>>
>>
>> Host: Greg Shakhnarovich <greg at ttic.edu>
>>
>>
>>
>> For more information on the colloquium series or to subscribe to the
>> mailing list, please see http://www.ttic.edu/colloquium.php
>>
>>
>>
>> Mary C. Marre
>> Administrative Assistant
>> *Toyota Technological Institute*
>> *6045 S. Kenwood Avenue*
>> *Room 504*
>> *Chicago, IL  60637*
>> *p:(773) 834-1757 <(773)%20834-1757>*
>> *f: (773) 357-6970 <(773)%20357-6970>*
>> *mmarre at ttic.edu <mmarre at ttic.edu>*
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20170612/8b334850/attachment-0001.html>