UTCS Colloquium/AI: Antonio Torralba/Massachusetts Institute of Technology Object Recognition by Scene Alignment ACES 2.402 Friday May 23 2008 11:00 a.m.

Contact Name: 
Jenna Whitney
May 23, 2008 11:00am - 12:00pm

*This is a re-scheduled talk originally scheduled
for April 11 2008

There is a signup schedule for this event (UT EID required).

Type of Talk: UTCS Colloquium/AI

Speaker/Affiliation: Antonio

Torralba/Massachusetts Institute of Technology

Date/Time: Friday M

ay 23 2008 11:00 a.m.

Location: ACES 2.402

Host: Kristen

Talk Title: Object Recognition by Scene Alignment


alk Abstract:
Object detection and recognition is generally posed as a m

problem between the object representation and the image feature

(e.g. aligning pictorial cues shape correspondence constellations

of parts etc.) while rejecting the background features using an outlie

process. In this work we take a different approach: we formulate the

object detection problem as a problem of aligning elements of the

entire scene. The background instead of being treated as a set of

iers is used to guide the detection process. Our approach relies
on th

e observation that when we have a big enough database then
we can find

with high probability some images in the database very
close to a query
image as in similar scenes with similar objects
arranged in similar s

patial configurations. If the images in the retrieval
set are partially
labeled then we can transfer the knowledge of the
labeling to the que

ry image and the problem of object recognition
becomes a problem of al

igning scene regions. But can we find a
dataset large enough to cover

a large number of scene configurations?
Given an input image how do we
find a good retrieval set and finally
how we do transfer the labels
to the input image? We will use two
datasets; 1) the LabelMe dataset
which contains more than 10 000
labeled images with over 180 000 annot

ated objects. 2) The tiny
images dataset: A dataset of weakly labeled i

mages with more than
79 000 000 images. We use this database to perform

object and scene
classification examining performance over a range of s

emantic levels.

Work in collaboration with Rob Fergus Bryan Russell
Ce Liu and William
T. Freeman.

Additional information and link