Sparse-Coding-for-Face-Image-Retrieval
Sparse-Coding-for-Face-Image-Retrieval copied to clipboard
work in Advanced Topics in Multimedia Analysis and Indexing
Sparse-Coding-for-Face-Image-Retrieval
Data
LFW DATA Labeled Faces in the Wild (LFW) is a widely used benchmark for face verification. It contains 13233 images of 5749 different people. You can find more information on the data set at http://vis-www.cs.umass.edu/lfw/.
Usage
Download the dataset.
cd ./src
python3 face_image_retrieval.py [LFW pickle data] [LFW attribute txt]
Dependency
Python3
numpy
joblib
scipy
pandas
matplotlib
spams
Baseline result
Compute L2 distance of local binary pattern (LBP) and rank.
Query | 1st | 2nd | 3rd | 4th | 5th |
---|---|---|---|---|---|
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
Sparse coding approach
Compute Cosine similarity of Spare coding with identity information and rank.
Query | 1st | 2nd | 3rd | 4th | 5th |
---|---|---|---|---|---|
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
-
mAP of vanilla sparse coding
-
mAP of sparse coding with identity information
Reference
[1] Scalable Face Image Retrieval using Attribute-Enhanced Sparse Codewords, Chen et al., IEEE Trans. Multimedia 2013
[2] Semi-supervised face image retrieval using sparse coding with identity constraint, Chen et al., ACM MM 2011
[3] SPArse Modeling Software