Garment Texture Classification for Automated Product Suggestion
•Download as PPTX, PDF•
1 like•725 views
This document presents a proposed method for classifying garment textures by analyzing local texture descriptors. It begins with background on the growing e-commerce market for garments and accessories. It then discusses issues with existing approaches and the need for an automated solution. The proposed method involves preprocessing images, identifying uniform colors, computing texture descriptors, and classifying patterns using random forest. It provides experimental results testing various descriptors on dresses and shirts, achieving over 90% accuracy with completed LBP. Future plans include expanding the method to more categories and accessories.
1 of 23
More Related Content
Garment Texture Classification for Automated Product Suggestion
4. US Garment Accessories e-shopping Revenue in 2013: 44.7 Billion Dollars [1]
45% of Consumers Prefer Shopping for Clothes Online [2]
64% Consult A Fashion Retailer's Website Before Making a purchase [2]
[1] http://www.statista.com/statistics/278890/us-apparel-and-accessories-retail-e-commerce-revenue
[2] https://econsultancy.com/blog/7960-45-of-consumers-prefer-shopping-for-clothes-online3/9/2015 4
7. Texture Match
Color Not Match
Match Not Match
Clothes Matching for Blind and Color
Blind People [3] (2010)
[3] Y. Tian, S. Yuan, “Clothes Matching for Blind and Color Blind People”, Springer, Computers Helping
People with Special Needs, Lecture Notes in Computer Science, Volume 6180, 2010, pp 324-331
3/9/2015 7
8. Dress
Short Pants
Identify Category of Garment Accessories [4] (2014)
[4] M. Manfredi. C. Grana, S. Calderara, R. Cucchiara, “A complete system for garment segmentation and
color classification”, Springer Machine Vision and Applications May 2014, Volume 25, Issue 4, pp 955-969
3/9/2015 8
10. Contribution of this thesis:
Propose a Pioneer method for Garment Texture Classification
3/9/2015 10
11. Issues related to Garment Textures:
Full of
shadows and
wrinkles
Large intra-class
variations in
appearance,
orientation & scale
Designed
with complex
patterns and
varieties of
colors
3/9/2015 11
15. Identify Uniform Color using
Color Descriptor
Uniform
Color
Yes
No
Input Image
3/9/2015 15
16. Compute Texture Pattern Using Texture
Descriptors
• LBP of Sign Component (Basic LBP)
• LBP of Magnitude Component
• LBP of Center Pixel3/9/2015 16
17. Classify Using Random Forest Classifier
Tree1
Tree2
Treen
Vote1
Vote2
Vote3
Voting
3/9/2015 17
18. Dataset
• Publicly available at the website:
http://imagelab.ing.unimore.it/fashion_dataset.asp
• Men Shirts and Girls’ Dresses were separated
• Then images were manually categorized into three
classes including uniform color, stripe and print
Girls’ Dress Men Shirt
Uniform Color 2441 200
Stripe 173 200
Print 1142 200
3/9/2015 18
19. Accuracy of The Descriptors
Texture Descriptor Accuracy
LBP 84.82
Mean LBP 79.22
Humming LBP 82.03
LTP 76.06
Fuzzy LBP 87.43
Noise Resistance LBP 92.10
Completed LBP 93.87
3/9/2015 19
21. Result (Men Shirt)
Category Precision Recall
Uniform Color 0.94 0.85
Print 0.86 0.90
Stripe 0.86 0.93
3/9/2015 21
22. Future Plan
• Compare in more descriptor-classifier combinations
• Fit the method more categories
• Enhance for other garment accessories like shoes, bags
3/9/2015 22