Using Adaptive Appearance Models Bogusaw Rymut, Bogdan Kwolek Rzeszw - - PowerPoint PPT Presentation

using adaptive appearance models
SMART_READER_LITE
LIVE PREVIEW

Using Adaptive Appearance Models Bogusaw Rymut, Bogdan Kwolek Rzeszw - - PowerPoint PPT Presentation

GPU-Supported Object Tracking Using Adaptive Appearance Models Bogusaw Rymut, Bogdan Kwolek Rzeszw University of Technology This paper describes how Graphics Processor Unit can be effectively used to speed-up the tracking algorithm based on


slide-1
SLIDE 1

GPU-Supported Object Tracking Using Adaptive Appearance Models

ICCVG 2010

Bogusław Rymut, Bogdan Kwolek

This paper describes how Graphics Processor Unit can be effectively used to speed-up the tracking algorithm based on adaptive appearance models. The object tracking is done by particle swarm optimization algorithm. Experimental results show that the GPU implementation of the algorithm exhibits a more than 40-fold speed-up over the CPU implementation.

Rzeszów University of Technology

slide-2
SLIDE 2

2

Agenda

 The problem  CUDA programming model  Particle Swarm Optimization  Problem decomposition  Experiments

slide-3
SLIDE 3

3

The problem

 Appearance based object tracking is

time-consuming

 The tracking algorithm must run in

real-time

 GPU implementation of PSO algorithm  Real-time tracking using PSO and

GPU

 How to decompose algorithm on GPU

slide-4
SLIDE 4

4

Object appearance

 

3 , , , 1 1

, 1 2 3

K k i k i k k i i k

f m M I initial intensity i previous intensity slow changes

 

         

 

t-1 t Fintess function 1 K 1 K

I

slide-5
SLIDE 5

5

CPU vs. GPU

1. www.nvidia.com

SIMD Architecture

slide-6
SLIDE 6

6

CUDA programming model

 Highly Multithreaded Coprocessor  Small set of extensions to C language  Low level programming  Focus on parallel algorithms

slide-7
SLIDE 7

7

CUDA programming model

 High scalable heterogeneous system

CPU & GPU are separate devices with separate DRAMs

GPU uses and executes thousand of extremely light threads to achive high performance

GPU DEVICE CPU DEVICE

slide-8
SLIDE 8

8

Particle Swarm Optimization

 Stochastic optimization algorithm  The optimization is achieved via set of

particles

 Particles collaborate each other in

  • ptimization process
slide-9
SLIDE 9

9

Particle Swarm Optimization

slide-10
SLIDE 10

10

Particle Swarm Optimization

( ) ( ) ( ) ( ) ( ) 1 1, 2 2, ( ) ( ) ( )

( ) ( )

i i i i i j j j j j j j j i i i j j j

v v c r pbest x c r gbest x x x v        

slide-11
SLIDE 11

11

Particle Swarm Optimization

1.

Assign each particle a random position in the problem hyperspace

2.

Evaluate the fitness function and find local best value for each particle

3.

Find the particle that has the best fitness value

4.

Update the velocities and positions of all particles

5.

Repeat steps 2-4 until maximum number of iterations is not attained

6.

Update appearance model

slide-12
SLIDE 12

12

Particle Swarm Optimization

slide-13
SLIDE 13

13

Approach to algorithm decomposition

 Each part of the algorithm has

been implemented as kernel function.

 Every particle has been

implemented as thread block

slide-14
SLIDE 14

14

Approach to algorithm decomposition

slide-15
SLIDE 15

15

Data decomposition

slide-16
SLIDE 16

16

Optimization of data access

 Access to on GPU global memory is

bottleneck

 Correctly data alignments essential to

  • verall performance
slide-17
SLIDE 17

17

Experiments

 PC with Intel Core 2 Quad 2.66 GHz,

1GB RAM

 PC with nVidia GeForce 9800 GT

14 multiprocessors 1.5 GHz, 1024MB RAM

slide-18
SLIDE 18

18

Face tracking

Real time Slow motion

slide-19
SLIDE 19

19

Experimental results Computation time [ms]

CPU [ms] 9800 GT [ms] Speedup

#32, 5 it 30.6 1.4 x22.4 #64, 5 it 60.0 1.9 x31.5 #128, 5 it 117.9 3.4 x38.8 #256, 5 it 234,2 5.6 x41.5

slide-20
SLIDE 20

20

Conclusions

 GPU implementation of PSO algorithm

has been prepared

 Our GPU based implementation is 40

times faster than CPU implementation