Feature Normalization for Part-Based Image Classification in ICIP 2013
"Explore the importance of feature normalization in part-based image classification as presented in ICIP 2013. Understand global, separate, and hierarchical normalization techniques and their impact on classification results. Dive into spatial pyramid matching (SPM) and hierarchical part matching (HPM) concepts. Discover the key step of feature normalization before classifiers for improved results."
Uploaded on | 0 Views
Download Presentation
Please find below an Image/Link to download the presentation.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author. If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.
You are allowed to download the files provided on this website for personal or commercial use, subject to the condition that they are used lawfully. All files are the property of their respective owners.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author.
E N D
Presentation Transcript
IEEE ICIP 2013 Feature Normalization for Part-Based Image Classification Speaker: Lingxi Xie Authors: Lingxi Xie, Qi Tian, Bo Zhang State Key Laboratory of Intelligent Technology and Systems Department of Computer Science and Technology Tsinghua University http://www.tsinghua.edu.cn
Outline Image Classification The Part-Based Classification Models Feature Normalization Global Normalization Separate Normalization Hierarchical Normalization Experimental Results Conclusions 6/11/2025 ICIP 2013 - Oral Presentation 2
Outline Image Classification The Part-Based Classification Models Feature Normalization Global Normalization Separate Normalization Hierarchical Normalization Experimental Results Conclusions 6/11/2025 ICIP 2013 - Oral Presentation 3
Image Classification A basic task towards image understanding General vs. Fine-Grained 6/11/2025 ICIP 2013 - Oral Presentation 4
Outline Image Classification The Part-Based Classification Models Feature Normalization Global Normalization Separate Normalization Hierarchical Normalization Experimental Results Conclusions 6/11/2025 ICIP 2013 - Oral Presentation 5
Spatial Pyramid Matching (SPM) = = = = = Part 1 [Lazebnik, CVPR06] Part 2 Part 3 Part 4 Part 5 6/11/2025 ICIP 2013 - Oral Presentation 6
Hierarchical Part Matching (HPM) = = = = = Part 1 [Xie, ICCV13] Part 2 Part 3 Part 4 Part 5 6/11/2025 ICIP 2013 - Oral Presentation 7
Outline Image Classification The Part-Based Classification Models Feature Normalization Global Normalization Separate Normalization Hierarchical Normalization Experimental Results Conclusions 6/11/2025 ICIP 2013 - Oral Presentation 8
Feature Normalization A Key Step before Classifiers Data pre-processing. Equivalent the range and weight of input vectors. Significantly impact the classification results. Different Models, Different Normalization. Support Vector Machine (SVM) Na ve Bayes Classificer (NB) Hidden Markov Models (HMM) 6/11/2025 ICIP 2013 - Oral Presentation 9
Global Normalization Considering the Feature as a Whole. coefficient norm 6/11/2025 ICIP 2013 - Oral Presentation 10
Global Normalization head head body body black-footed albatross sooty albatross 6/11/2025 ICIP 2013 - Oral Presentation 11
Global Normalization 1 1 2 2 3 3 4 4 5 5 6 6 1 1 2 2 3 3 4 4 5 5 6 6 head head body body ALL ALL head head body body ALL ALL not balanced 6/11/2025 ICIP 2013 - Oral Presentation 12
Global Normalization Small Parts have Low Feature Weights. 5 FAIR? 1 2 3 4 6 1 2 3 4 5 6 head body ALL head body ALL 6/11/2025 ICIP 2013 - Oral Presentation 13
Global Normalization However, Small Parts are also Important. groove billed ani common raven red winged blackbird rusty blackbird beak wing 6/11/2025 ICIP 2013 - Oral Presentation 14
Separate Normalization Normalizing each Part Individually part-wise vector coefficient 6/11/2025 ICIP 2013 - Oral Presentation 15
Separate Normalization 1 1 2 2 3 3 4 4 5 5 6 6 1 1 2 2 3 3 4 4 5 5 6 6 head head body body ALL ALL head head body body ALL ALL Small Parts are Given Equal Weights! 6/11/2025 ICIP 2013 - Oral Presentation 16
Separate Normalization Different-Level Parts have Same Weight. 1 2 3 4 5 FAIR? 6 1 2 3 4 5 6 head body ALL head body ALL 6/11/2025 ICIP 2013 - Oral Presentation 17
Separate Normalization Examples of Hierarchical Parts in Birds Dataset. head = beak + eyes + crown + forehead neck = nape + throat body = breast + back + belly + wings ALL = head + body + tail + legs Some Observations: High-level parts contain more information. High-level parts are less likely to be missing. 6/11/2025 ICIP 2013 - Oral Presentation 18
Hierarchical Normalization Assigning more Weights on High-Level Parts hierarchical contribution part-wise weight part-wise coefficient 6/11/2025 ICIP 2013 - Oral Presentation 19
Hierarchical Normalization 1 1 2 2 3 3 4 4 5 5 6 6 1 1 2 2 3 3 4 4 5 5 6 6 head head body body ALL ALL head head body body ALL ALL High-Level Parts are Enhanced! 6/11/2025 ICIP 2013 - Oral Presentation 20
Outline Image Classification The Part-Based Classification Models Feature Normalization Global Normalization Separate Normalization Hierarchical Normalization Experimental Results Conclusions 6/11/2025 ICIP 2013 - Oral Presentation 21
The Caltech101 Dataset General Object Recognition [Fei-Fei, CVIU07] 102 classes (one background category) 9144 images Models SPM [Lazebnik, CVPR06] + LLC [Wang, CVPR10] SPM [Lazebnik, CVPR06] + GPP [Xie, MM12] SPM [Lazebnik, CVPR06] + EdgeGPP [Xie, MM12] 16 Basic Parts + 5 High-Level Parts 6/11/2025 ICIP 2013 - Oral Presentation 22
The Caltech101 Dataset SPM+LLC 73.14 73.91 74.41 73.25 71.99 73.68 73.39 72.71 74.31 73.89 SPM+GPP 76.35 76.26 77.03 76.47 75.20 75.47 76.43 75.88 76.86 76.55 SPM+EdgeGPP 80.78 80.86 82.45 80.89 78.05 81.24 80.93 80.40 83.19 81.37 No Normalization Global L1-norm Global L2-norm Global Li-norm Separate L1-norm Separate L2-norm Separate Li-norm Hierarchical L1-norm Hierarchical L2-norm Hierarchical Li-norm 6/11/2025 ICIP 2013 - Oral Presentation 23
The CUB-200-2011 Dataset Fine-Grained Bird Classification [Wah, TR11] 200 species of birds 11788 images Models Parts [Xie, ICCV13] + LLC [Wang, CVPR10] HPM [Xie, ICCV13] + LLC [Wang, CVPR10] HPM [Xie, ICCV13] + GPP [Xie, MM12] 15 Basic Parts + 6 High-Level Parts 6/11/2025 ICIP 2013 - Oral Presentation 24
The CUB-200-2011 Dataset Parts+LLC 25.58 27.61 27.06 25.04 24.58 30.93 27.73 HPM+LLC 27.55 29.85 29.30 27.41 26.94 32.75 29.67 28.12 32.89 29.45 HPM+GPP 30.67 33.22 32.96 30.71 31.84 35.98 31.92 33.85 36.48 32.08 No Normalization Global L1-norm Global L2-norm Global Li-norm Separate L1-norm Separate L2-norm Separate Li-norm Hierarchical L1-norm Hierarchical L2-norm Hierarchical Li-norm - - - 6/11/2025 ICIP 2013 - Oral Presentation 25
Discussions The Performance of Our Model SPM: only comparable with the original model. HPM: significantly better! Why? Both Assumptions Sound? There exist more semantics in HPM! 6/11/2025 ICIP 2013 - Oral Presentation 26
Outline Image Classification The Part-Based Classification Models Feature Normalization Global Normalization Separate Normalization Hierarchical Normalization Experimental Results Conclusions 6/11/2025 ICIP 2013 - Oral Presentation 27
Conclusions Feature Normalization An important issue in image representation. In Part-Based Classification Models Instructive to consider each part separately. 3 Normalization Strategies Global Normalization Separate Normalization Hierarchical Normalization Easy to Implement! 6/11/2025 ICIP 2013 - Oral Presentation 28
Thank you! Questions please? 6/11/2025 ICIP 2013 - Oral Presentation 29
