Download presentation
Presentation is loading. Please wait.
1
1 Integrating 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.
2
2 Outline Introduction Background Log-based Relevance Feedback Coupled Support Vector Machine Support Vector Machine Formulation Alternating Optimization A Practical Algorithm Experimental Results Conclusion
3
3 Introduction 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
4
4 Introduction 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.
5
5 Introduction Motivation Can user feedback log be used to improve the regular relevance feedback? Relevance Feedback User Feedback Log ? Problem
6
6 Background Log-based Relevance Feedback (LRF) Relevance Matrix: R RF round / Log session: N l images are marked Elements: relevant (1), irrelevant (-1), unknown (0) Log Sessions Image samples 1 1 0 1 11 1 1 0
7
7 Background Learning Problem for LRF Low-level image content: User feedback log: Multi-Modal Learning Problem
8
8 Coupled 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
9
9 A Straightforward Solution For the image content modality: w T x For the user feedback log modality: u T r Coupled Support Vector Machine
10
10 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. Coupled Support Vector Machine
11
11 Formulation: Coupled SVM Coupled Support Vector Machine
12
12 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 ’ Coupled Support Vector Machine
13
13 Alternating Optimization Fix Y ’, the primal optimization is equivalent to solving the two optimization subproblems: Coupled Support Vector Machine
14
14 Alternating Optimization (AO) By introducing non-negative Lagrange multipliers, the above two subproblems can be solved Coupled Support Vector Machine
15
15 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: Coupled Support Vector Machine
16
16 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. Coupled Support Vector Machine
17
17 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 Coupled Support Vector Machine
18
18 A Practical Algorithm (cont ’ d) Coupled Support Vector Machine
19
19 A Practical Algorithm (cont ’ d) Coupled Support Vector Machine
20
20 Experimental 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
21
21 Experimental Results (cont ’ d) 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
22
22 Experimental Results (cont ’ d) 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
23
23 Experimental Results (cont ’ d) CBIR GUI for collecting feedback data
24
24 Experimental Results (cont ’ d) 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
25
25 Experimental Results (cont ’ d) Performance Evaluation: on 20-Category Dataset
26
26 Experimental Results (cont ’ d) Performance Evaluation: on 50-Category Dataset
27
27 Experimental Results (cont ’ d)
28
28 Experimental Results (cont ’ d)
29
29 Conclusion 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.
30
30 Q&A
31
31 References 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.
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.