Title: Integrating User Feedback Log into Relevance Feedback by Coupled SVM for Content-Based Image Retrieval
1Integrating User Feedback Log into Relevance
Feedback by Coupled SVM for Content-Based Image
Retrieval
- 9-April, 2005
- Steven C. H. Hoi , Michael R. Lyu , Rong Jin
- Department of Computer Science Engineering
- The Chinese University of Hong Kong
- Shatin, N.T., Hong Kong SAR
- Department of Computer Science and Engineering
- Michigan State University
- East Lansing, MI 48824, USA
- The 1st IEEE EMMA Workshop
- in conjunction with 21st IEEE ICDE, Japan, April,
2005.
2Outline
- Introduction
- Background
- Log-based Relevance Feedback
- Coupled Support Vector Machine
- Support Vector Machine
- Formulation
- Alternating Optimization
- A Practical Algorithm
- Experimental Results
- Conclusion
3Introduction
- Content-based Image Retrieval (CBIR)
- An important component in visual information
retrieval - QBE query-by-example based on low-level visual
features - Semantic gap low-level features, high-level
concepts
QBE
4Introduction
- Relevance Feedback (RF)
- A powerful tool to attack the semantic gap
problem - Interactive mechanism to solicit users feedbacks
- Boost the retrieval performance of CBIR greatly
- Many existing techniques already
- Problems
- Regular relevance feedback needs too many rounds
of interactions for achieving satisfactory
results.
5Introduction
Relevance Feedback
?
User Feedback Log
Can user feedback log be used to improve the
regular relevance feedback?
Problem
6Background
- Log-based Relevance Feedback (LRF)
- Relevance Matrix R
- RF round / Log session Nl images are marked
- Elements relevant (1), irrelevant (-1), unknown
(0)
Image samples
-1
-1
1
1
1
-1
-1
0
1
-1
1
-1
1
-1
-1
-1
-1
-1
0
-1
-1
1
Log Sessions
7Background
- Learning Problem for LRF
- Low-level image content
- User feedback log
- Multi-Modal Learning Problem
8Coupled Support Vector Machine
- Motivation
- How to attack the learning problem on the two
modalities? - Low-level Image content X
- User relevance feedback log R
- Support Vector Machines superior classification
performance - A Straightforward Solution
- Learn an SVM classifier on each modality
respectively - For image content X, we learn an optimal
weighting vector w - For log content R, we learn an optimal weighting
vector u - Combine their results together linearly
9Coupled Support Vector Machine
- A Straightforward Solution
- For the image content modality wTx
- For the user feedback log modality uTr
10Coupled Support Vector Machine
- Disadvantages of the straightforward solution
- Linear combination
- Modality Consistence
- Our better solution Coupled SVM
- Learn the two modalities in a unified formulation
- Enforce the prediction on the two types of
information to be consistent.
11Coupled Support Vector Machine
12Coupled Support Vector Machine
- Optimization of Coupled SVM
- Hard to be solved directly
- Alternating Optimization (AO)
- AO two-step optimization
- Fix Y, try to find (u, b_u), and (w, b_w)
- Fix (u, b_u) and (w, b_w), try to find Y
13Coupled Support Vector Machine
- Alternating Optimization
- Fix Y, the primal optimization is equivalent to
solving the two optimization subproblems
14Coupled Support Vector Machine
- Alternating Optimization (AO)
- By introducing non-negative Lagrange multipliers,
the above two subproblems can be solved
15Coupled Support Vector Machine
- Alternating Optimization (AO)
- After solving (u, b_u) and (w, b_w), fixing them,
the optimal Y can be found to fit the data as
follows
16Coupled Support Vector Machine
- Summary of AO procedure
- 1) Beginning with a small value of
- 2) Performing the two-step AO procedure
- 3) Repeating 2) by increasing until it
achieves the setting threshold - Comments on the Coupled SVM
- Can be a general approach for multi-modal
learning problems - Need to investigate the convergence issue of
Alternating Optimization - Need to study better methods for solving the
optimization problem - Require to take some practical considerations
when fitting for specific problems.
17Coupled Support Vector Machine
- A Practical Algorithm
- Practical considerations
- Cannot engage all unlabeled samples due to
response requirement for relevance feedback - Strategy for choosing unlabeled samples
- Closest to the decision boundary of SVM most
informative according to active learning - Closest to the labeled samples to avoid too much
effort in learning the label information - Introducing a parameter to control the
error for label correction to avoid overlarge
change in the labeled set
18Coupled Support Vector Machine
- A Practical Algorithm (contd)
19Coupled Support Vector Machine
- A Practical Algorithm (contd)
20Experimental Results
- Dataset
- Images selected from COREL image CDs
- Two ground-truth datasets
- 20-Category each category contains 100 images,
totally 2,000 - 50-Category each category contains 100 images,
totally 5,000
21Experimental Results (contd)
- Low-level Image Representation
- Color Moment
- 9-dimension
- Edge Direction Histogram
- 18-dimension
- Canny detector, 18 bins of 20 degrees each
- Wavelet-based texture
- 9-dimension
- Daubechies-4 wavelet, 3-level DWT
- Entropies of 9 subimages are generated for the
texture feature
22Experimental Results (contd)
- Collection of User Log Data
- Log format
- A log session (LS) corresponds a relevance
feedback round - Each log session contains 20 images labeled by
users - Log data
- On 20-Category 161 log sessions
- On 50-Category 184 log sessions
23Experimental Results (contd)
- CBIR GUI for collecting feedback data
24Experimental Results (contd)
- Performance Evaluation
- Measurement Metric
- Average Precision relevant images /
returned images - Experimental Setting
- 100 queries
- 20 initially labeled images
- SVM RBF kernel, parameters set via training data
- Comparison Schemes
- RF-SVM
- traditional relevance feedback by SVM
- LRF-2SVM
- log-based relevance feedback by learning two SVMs
respectively - LRF-CSVM
- log-based relevance feedback by Coupled SVM
25Experimental Results (contd)
- Performance Evaluation on 20-Category Dataset
26Experimental Results (contd)
- Performance Evaluation on 50-Category Dataset
27Experimental Results (contd)
28Experimental Results (contd)
29Conclusion
- A log-based relevance feedback scheme was studied
by integrating user feedback log into the content
learning of low-level visual features in
content-based image retrieval. - A general multimodal learning technique, i.e.
Coupled Support Vector Machine, was proposed for
studying the data with multiple modalities. - A practical algorithm by Coupled SVM was
presented to attack the log-based relevance
feedback problem in CBIR. - Experimental results show our proposed scheme is
effective for the log-based relevance feedback
problem.
30QA
31References
- Chu-Hong Hoi and Michael R. Lyu, A Novel
Log-based Relevance Feedback Technique in
Content-based Image Retrieval, in Proc. ACM
Multimedia, New York, USA, 10-16 October, pp.
24-31, 2004 - S. Tong and E. Chang. Support vector machine
active learning for image retrieval. In Proc. ACM
Multimedia, pages 107--118, 2001.