Download presentation
Presentation is loading. Please wait.
1
TENNIS STROKE DETECTION
CS365 Mentor: Prof. Amitabh Mukherjee Khushdeep Singh (10351) Aakash Verma (10002)
2
Applications of human activity recognition
Khushdeep Singh and Aakash Verma
3
OUR task can be divided into the following subtasks
Player Tracking Optical Flow Analysis Using Motion Descriptors Adaptive Boosting Khushdeep Singh and Aakash Verma
4
Player Tracking Using Particle Filter
5
What must be specified:
Prior Distribution: p(x0) - Describes initial distribution of object states Transition Model: p(xt | xt-1) - Specifies how objects move between frames - A simple model: sample next state from a Gaussian window around current state - We used second order auto regressive model. xt = Axt-1 + Bxt-2 +wt Observation Model: p( yt | xt ) - Color Histogram Object Tracking and Particle Filtering by Rob Hess [2006] Khushdeep Singh and Aakash Verma
6
EXAMPLE OF PLAYER TRACKING
Khushdeep Singh and Aakash Verma
7
OPTICAL FLOW Using KL ALGORITHM 11-07-2019
Khushdeep Singh and Aakash Verma
8
OPTICAL FLOW estimation
SPARSE OPTICAL FLOW Computed only at a subset of image points. Quicker but less accurate results. Example: Kanade-Lucas Algorithm DENSE OPTICAL FLOW Computed at each image pixel. Slower but better results. Example: Farneback Algorithm Khushdeep Singh and Aakash Verma
9
EXAMPLE OF SPARSE OPTICAL FLOW ESTIMATION
Khushdeep Singh and Aakash Verma
10
ITERATIVE KL OPTICAL FLOW COMPUTATION
Khushdeep Singh and Aakash Verma
11
Feature DETECTION USING VOILA JONES ALGORITHM 11-07-2019
Khushdeep Singh and Aakash Verma
12
VOILA JONES algorithm Efficient Visual Event Detection using Volumetric Features by Yan Ke, Rahul Suthankar, Martial Herbert [ICCV’05] Khushdeep Singh and Aakash Verma
13
ADAPTIVE BOOSTING Meta machine learning algorithm 11-07-2019
Khushdeep Singh and Aakash Verma
14
Lecture on AdaBoost by Jan Sochman, Jiri Matas
Lecture on AdaBoost by Jan Sochman, Jiri Matas Khushdeep Singh and Aakash Verma
15
For the given example Khushdeep Singh and Aakash Verma
16
Comparison of adaboost with other methods
BoosTexter: A boosting-based system for text categorization by Robert E. Schapire and Yoram Singer. Khushdeep Singh and Aakash Verma
17
Adaboost pseudocode Khushdeep Singh and Aakash Verma
18
Some initial results using only 1000 weak classifiers
No. of Training Examples No. of Features Misses False Alarms 200 2,00,000 3% 60% 19.58% 30% 500 1,00,000 12% 58% * TRAINED ON THE KTH DATASET Khushdeep Singh and Aakash Verma
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.