RECENT
RESEARCH TOPICS 
Deep Manifold
Distance Learning

Image distances are
efficiently estimated on a manifold of
select image representatives, using a
closedform convergence solution of the
Random Walk, for image retrieval and
clustering.


Learning to
Learn SecondOrder BackPropagation Using
LSTMs

We derive an efficient
backpropagation for computing both
gradients and second derivatives of a
CNN’s loss. These are then input to an
LSTM for predicting optimal updates of
CNN parameters.


Temporal
Deformable Residual Networks for Action
Segmentation

Our TDRN computes deformable
temporal convolutions in two parallel
streams at short and longrange scales
for action segmentation.


Boundary Flow

We specify a Siamese network
for joint estimation of object
boundaries and their motion in videos.
Boundary flow is an important midlevel
visual cue as boundaries and their flow
more explicitly indicate objects'
motions and interactions than common
dense optical flow.


Recognition as
HSnet Search for Informative Image Parts

We specify a sequential
search for informative object parts over
a deep feature map in the image toward
finegrained recognition.


Combining
BottomUp, TopDown, and Smoothness Cues for
Weakly Supervised Segmentation

We fuse three processes
toward semantic segmentation  namely,
(i) Bottomup computation of neural
activations in a CNN for the imagelevel
class prediction; (ii) Topdown
estimation of attention maps of the
CNN's activations given the predicted
objects; and (iii) Lateral
attentionmessage passing among
neighboring neurons at the same CNN
layer.


BudgetAware
Semantic Video Segmentation

We learn a budgetaware
policy to optimally select a small
subset of frames for pixelwise labeling
by a CNN within a given time limit, and
then efficiently interpolate the
obtained segmentations to yet
unprocessed frames.


Unsupervised
Video Summarization

We use a deep generative
adversarial network to select a sparse
subset of video frames which optimally
represent the input video.


ConfidenceEnergy
Recurrent Network for Activity Recognition

We use a twolevel hierarchy
of Long ShortTerm Memory (LSTM)
networks for predicting individual
actions, interactions, and group
activities. When estimating the energy
of our predictions, we additionally
compute pvalues of the solutions, and
in this way choose the most confident
energy minima.


Affordance
Segmentation in Images

Pixels of indoor scenes are
labeled with five affordance types:
walkable, sittable, lyable, reachable,
and movable using a cascade of
convolutional neural networks (CNNs).
CNNs first extract midlevel visual
cues, including: depth map, surface
normals, and coarse semantic scene
segmentation, and then combine them
toward affordance segmentation.


Recurrent
Temporal Deep Field for Semantic Video
Labeling

A new deep architecture,
called Recurrent Temporal Deep Field
(RTDF), is specified for semantic video
labeling. RTDF is a conditional random
field that combines a deconvolution
neural network and a recurrent temporal
restricted Boltzmann machine.


Leveraging
HumanSkeleton Sequences for Action
Recognition in Videos

A largescale action
recognition in videos can be greatly
improved by providing an additional
modality in training  3D
humanskeleton sequences.


Monocular
Depth Estimation Using Neural Regression
Forest

A new deep architecture 
neural regression forest (NRF)  is
used for depth estimation from a single
image. NRF consists of regression trees,
whose each node corresponds to a
convoutional neural network (CNN).


Budgeted
Semantic Video Segmentation

Our goal is to accurately
label all pixels in the video within a
specified time budget. We formulate a
new budgeted inference for a CRF that
intelligently selects the most useful
subsets of features to be extracted from
subsets of supervoxels in the video
within the time budget.


Continuous
Facial Behavior Estimation

A Doubly Sparse Relevance
Vector Machine is specified to enforce
double sparsity in the joint selection
of the most relevant training examples,
and the most important kernels
associated with facial parts for
continuous estimation of facial
behavior.


Semantic
Segmentation of RGBD Images with Mutex
Constraints

Pixels are assigned class
labels, such that the labeling strictly
satisfies all mutual exclusion (mutex)
constraints, and thus does not violate
commonsense physics laws. The mutex
constraints considered include: object
cooccurrence, relative height and
support relationships.


TreeCut for
Probabilistic Image Segmentation

Given an image and its region
tree, image segmentation is formalized
as sampling cuts in the tree using
dynamic programming. Our treecut model
can be tuned to sample segmentations at
a particular scale of interest, and thus
conduct a scalespecific evaluation.


Joint
Inference of Groups, Events and Human roles in
Aerial Videos

We parse lowresolution
aerial videos of large spatial areas, in
terms of detecting events, and grouping
and assigning roles to people engaged in
the events. The challenges arising from
low resolution and topdown views are
addressed by conducting joint inference
of these tasks.


Latent Trees
for Estimating Intensity of Facial Action
Units

A new generative latent tree
(LT) model is specified for FAU
intensity estimation. Our new structure
learning iteratively builds LT so as to
maximize the likelihood and minimize the
model complexity. We derive closedform
expressions of posterior marginals for
all variables in LT, and specify an
efficient bottomup/topdown inference.


HCSearch for
Structured Prediction in Computer Vision

Semantic scene segmentation
and monocular depth estimation are
formulated as a search problem. The
search space is defined by probabilistic
sampling of plausible image
segmentations.


Person Count
Localization in Videos

Given a video, we output for
each frame a set of: 1) Detections
optimally covering both isolated
individuals and cluttered groups of
people; and 2) Counts of people inside
these detections. This problem is a
middleground between framelevel person
counting, which does not localize
counts, and person detection aimed at
perfectly localizing people with
countone detections.


Monocular
Extraction of 2.1D Sketch Using Constrained
Convex Optimization

We partition the image into
regions, and estimate their depth
ordering in the scene. This is cast as a
constrained convex optimization problem,
and solved within the optimization
transfer framework. Our new optimization
transfer admits a closedform expression
of the duality gap, and thus allows
explicit computation of the achieved
accuracy.


HiRF:
Hierarchical Random Field for Collective
Activity Recognition in Videos

We formulate Hierarchical
Random Field (HiRF) for activity
recognition. HiRF establishes strictly
hierarchical links between all
variables, discarding
the common lateral temporal connections.
This enables an efficient
bottomup/topdown inference.


MultiObject
Tracking via Constrained Sequential Labeling

Constrained sequential
labeling (CSL) assigns object
identifiers to supervoxels, while
respecting domain constraints. CSL is
wellsuited for simultaneous labeling
and fixing noisy merges and splits of
our midlevel features, which cannot be
handled in a principled manner by
traditional network flow approaches. CSL
is efficient due to contraint
propagation.


Scene Labeling
Using Beam Search Under Mutex Constraints

We cast scene labeling as
quadratic program (QP) with mutual
exclusion (mutex) constraints on class
label assignments. The QP is solved
efficiently using beam search,
which explicitly accounts for spatial
extents of objects, and guarantees that
all mutex constraints from domain
knowledge are satisfied.


Play Type
Recognition in RealWorld Football Video

Given a video sequence of
plays of a football game, we integrate
responses of the playlevel detectors
with global gamelevel reasoning to
overcome huge variations in camera
viewpoint, motion, and distance from the
field, as well as amateur camerawork
quality.


Inferring Dark
Matter and Dark Energy from Videos

Functional objects do not have
discriminative appearance and shape, but
can be viewed as "dark matter",
emanating "dark energy" that affects
people’s trajectories in the video. For
localizing functional objects, we
analyze noisy behavior of people in the
scene using agentbased, probabilistic
Lagrangian mechanics.


Latent
Multitask Learning for ViewInvariant Action
Recognition

When each viewpoint of a given
set of action classes is specified as a
learning task then multitask learning
appears suitable for achieving view
invariance in recognition. We extend the
standard multitask learning to allow
identifying: (1) latent groupings of
action views (i.e., tasks), and (2)
discriminative action parts, along with
joint learning of all tasks.


Monte Carlo
Tree Search for Scheduling Activity
Recognition

Querying an activity in a long
video footage may require running a
multitude of detectors, and tracking
their detections. We use Monte Carlo
Tree Search to optimally schedule a
sequence of detectors and trackers to be
run, and where they should be applied in
the spacetime volume.


SLEDGE:
Sequential Labeling of Image Edges for
Boundary Detection

We sequentially label image
edges as "on" or "off" object
boundaries. A visited edge is labeled as
boundary based on evidence of its
perceptual grouping with already
identified boundaries. We use both local
Gestalt cues, and the global Helmholtz
principle of nonaccidental grouping.
Image edges are extracted with our new
detector that finds salient pixel
sequences which separate distinct
textures within the image.


Hough Forest
Random Field for Object Recognition and
Segmentation

We combine Hough forest (HF)
and conditional random field (CRF) into
HFRF to assign labels of object classes
to image regions. HF captures intrinsic
and contextual properties of objects as
class histograms in the leaf nodes. This
evidence is used in CRF inference for
nonparametric density estimation of the
posteriors. Theoretical error bounds of
HF and HFRF applied to a twoclass
object detection and segmentation are
also presented.
Personal use of
this material is permitted.
However, permission to
reprint/republish this material
for advertising or promotional
purposes or for creating new
collective works for resale or
redistribution to servers or
lists, or to reuse any
copyrighted component of this
work in other works must be
obtained from the IEEE.


CostSensitive
Topdown/Bottomup Inference for Multiscale
Activity Recognition

While prior work typically
addresses activity recognition at a
single scale, we jointly model group
activities, individual actions, and
participating objects with an ANDOR
graph, and exploit its hierarchical
structure for efficient inference. An
exploreexploit strategy is used to
adaptively zoomin or zoomout for
costsensitive inference.


Human Actions
as Stochastic Kronecker Graphs

A human activity can be
viewed as a spacetime repetition of
activity primitives. Given a set of
primitives, and an affinity matrix of
their probabilistic grouping, we
formulate that a video of the activity
is probabilistically generated by a
sequence of the Kronecker products of
the affinity matrix.


SumProduct
Networks for Modeling Activities with
Stochastic Structure

Activities with stochastic
structure are characterized by variable
spacetime arrangements of
subactivities, and may be conducted by a
variable number of actors. We use
sumproduct networks (SPN) to model such
activities. The products are aimed at
encoding particular configurations of
activity parts, and the sums serve to
capture their alternative
configurations. A new Volleyball dataset
is collected and annotated.


Learning
Spatiotemporal Graphs of Human Activities

Given a set of spatiotemporal
graphs, we learn their model graph, and
pdf's associated with nodes and edges of
the model. The model graph adaptively
learns from data relevant video segments
and their spatiotemporal relations. We
present a novel weightedleastsquares
formulation of learning a structural
archetype of graphs. The model is used
for video parsing.


From Contours
to 3D Object Detection and Pose Estimation

We address viewinvariant
object detection and pose estimation
using contours as basic object features.
A topdown feedback from inference warps
the image, so the bottomup extraction
of contours could better collectively
summarize relevant visual information
and match our 3D object model, under
arbitrary nonrigid shape deformations
and affine projection.


A Chains Model
for Localizing Participants of Group
Activities in Videos

We address recognition of
group activities in a given video,
localization of video parts where these
activities occur, and detection of
actors involved in them. A new
generative chains model is formulated to
organize a large number of video
features in an ensemble of chains,
starting and ending at the end points of
the time interval occupied by the target
activity.


Scene Shape
from Texture of Objects

We estimate the 3D shape of a
scene from texture arising from a
spatial repetition of objects in the
image. Unlike existing work, our
monocular estimation does not use domain
knowledge about the layout of common
scene surfaces. We also show that
reasoning about texture of objects in
the scene improves object detection.


Multiobject
Tracking as Maximum Weight Independent Set

We prove that the data
association problem  the core of
tracking  can be formulated as finding
the maximumweight independent set
(MWIS) of nonadjacent tracklets in a
graph. We present a new, polynomialtime
MWIS algorithm, and prove that it
converges to an optimum.


Probabilistic
Event Logic for IntervalBased Event
Recognition

We introduce probabilistic
event logic (PEL) for representing both
hard and soft temporal constraints among
events. A PEL knowledge base consists of
confidenceweighted formulas from a
temporal event logic, and specifies a
joint distribution over the occurrence
time intervals of all events. Our MAP
inference for PEL addresses the
scalability issue of reasoning about all
time intervals in video, by leveraging
the spanninginterval data structure. A
spanning interval compactly represents
entire sets of time intervals without
enumerating them.


(RF)^2 
Random Forest Random Field

We combine random forest (RF)
and conditional random field (CRF) to
address multiclass object recognition
and segmentation. Inference of (RF)^2
uses MetropolisHastings jumps which
depend on two ratios of the proposal and
posterior distributions. Our key idea is
to directly learn these ratios using RF.


Segmentation
as MaximumWeight Independent Set

Given an image, and an
ensemble of its distinct lowlevel
segmentations, we identify visually
"meaningful" segments in the ensemble.
This is formalized as the maximumweight
independent set (MWIS) problem. We
formulate a new MWIS iterative
algorithm, where each iteration solves a
Taylor expansion of the MWIS objective
function in the discrete domain.


Activities as
Time Series of Human Postures

We show that certain human
actions can be represented by short time
series of codewords. The codewords
represent still snapshots of humanbody
parts in their discriminative postures,
and objects that people interact with
while performing the activity. This
carries many advantages for developing a
robust, efficient, and scalable activity
recognition system.


From a Set of
Shapes to Object Discovery

We show that shape is
expressive and discriminative enough to
provide robust object discovery in the
midst of background clutter. We build a
graph that captures spatial layouts of
edges extracted from a set of images,
and conduct its multicoloring by a new
coordinate ascent SwendsenWang cut. The
resulting clusters of edges delineate
the boundaries of distinct objects
discovered in the image set.


Monocular
Extraction of 2.1D Sketch

Given a segmentation and
Tjunctions of an image, we estimate the
depth layers of the scene. The
estimation is formalized as a quadratic
optimization so the resulting 2.1D
sketch is smooth in all image areas
except on region boundaries.


Video Painting
with SpaceTime Varying Style Parameters

An input video is rendered by
applying a distinct painting style to
each spatiotemporal tube, corresponding
to a moving object in the video.
Spatiotemporal segmentation allows
the user a control to vary
painting styles in 2D space and time,
and thus convey rich semantic
content, e.g., emotions, illusion,
chaos, etc.


Toward Optimal
Feature Selection through Local Learning

Given data with a huge number
of irrelevant features (> 10 ^{6}),
select features relevant to data
classification. We decompose a nonlinear
problem into a set of locally linear
ones, and then globally learn feature
relevance within the large margin
framework.


Video Object
Segmentation by Tracking Regions

Given an arbitrary video,
segment all moving and static objects
present. We transitively match contours
of image regions across the frames such
that the resulting tracks are locally
smooth.


Texelbased
Texture Segmentation

Given an arbitrary image,
discover and segment all distinct
texture subimages. We use the meanshift
to simultaneously estimate the pdf of
texel appearance and the pdf of texel
placement.


Matching
Hierarchies of Deformable Shapes

Shapes are represented by
graphs whose nodes correspond to shape
parts, and edges capture their neighbor
and partof interactions. Shape matching
is formulated as finding the subgraph
isomorphism that minimizes a quadratic
cost.


DictionaryFree
Categorization
Using Evidence Trees

How to categorize images
showing very similar object categories?
We mathematically prove that it is
better to use class evidence accumulated
from all image features than to use a
majority voting of class decisions made
on each individual feature.


Scaleinvariant
Regionbased
Hierarchical Image Matching

Find correspondences between
similar objects in images captured under
large variations in scale. Scale
invariance is achieved by decoupling the
scales of objects from those of scenes,
and by downweighting the contributions
of fineresolution details to matching.


Learning
Subcategory Relevances for Category
Recognition

Detections of distinct object
categories provide different degrees of
evidence for recognition of more
complex, parent categories. This is
estimated using local learning.


Connected
Segmentation Tree
 A Joint
Representation of Region Layout and Hierarchy


CST is a hierarchy of region
adjacency graphs. The CST model of an
object category is learned by
simultaneously searching for both the
most salient regions, and the most
salient containment and neighbor
relationships of regions across training
images.


Extracting
Texels in 2.1D Natural Textures

Given an image of 2.1D
texture, learn without any supervision a
generative model of the entire
(unoccluded) texel. Learning involves
concurrent estimation of the
texelsubtexel structure, and the pdf's
of each texel part from only partially
visible texels in the image.


Taxonomy of
Categories Present in Arbitrary Images

Given an
arbitrary (unlabeled) image set, learn
the models of all visual categories
present, and their intercategory
relationships, i.e., their taxonomy. The
taxonomy recursively defines categories
as spatial configurations of (simpler)
subcategories each of which may be
shared by many categories.



The hoofed animals dataset
contains very similar categories that
share a number of similar parts. Each
image may contain multiple instances of
multiple categories. Animals are
articulated, nonrigid objects,
appearing at different scales amidst
clutter, and may be partially occluded.



The images show homogeneous,
frontally viewed, natural, 2.1D
textures, where: (1) Texels are only
statistically similar to each other; (2)
Texel placement is random; (3)
Repetition of subtexels define a finer
grain texture coexisting with the main
texture; (4) Due to texel overlap, texel
contours form complex patterns (e.g.,
several edges meet at one point), and
overlapping texels have low contrasts,
all of which makes texel segmentation
difficult.


Unsupervised
Category Modeling, Recognition and
Segmentation

Given a set of
images containing frequent occurrences
of an unknown visual category, learn
geometric, photometric and topological
properties of regions defining the
category. Learning is unsupervised,
because the target category is not
defined by the user, and whether and
where any instances of the category
appear in a specific image is not known.

