=======================================================================================
This matlab source code comes from http://www.unc.edu/~yunchao/itq.htm.
I have modified it and implement drawing the Precision Recall Curves using
different methods such as ITQ/LSH/RR/SKLSH, and different binary bit length
PR_Curves using LSH. A Retri_Images_Demo is also provided to show the image
retrieval result using ITQ.
Modified by Willard, and you can see the detains from my website is:
========================================================================================
- ITQ.m: implements the Iterative Quantization method
- compressITQ: convert large-scale data to binary matrix
- main.m: demo code computes recall precision, also implements some other baseline methods
- RF_train.m: estimate some parameter for SKLSH
- RF_compress.m: perform SKLSH binary embedding
- cca.m performs a canonical correlation analysis for supervised cases
- Demo_PR.m: This is a geometric illustration of Draw the Recall Precision Curve
- DemoLSH_PR.m: This is a geometric illustration of LSH recall and precision using
different code length - Retri_Images_Demo.m: This is a PCA-ITQ demo showing the retrieval sample
========================================================================================
- before ITQ, the data must have be centered
- we found 20-50 iterations are usually enough for ITQ
- I modified it on a 16G memory computer, 4G is failed for the dataset is large
========================================================================================
Demo results
Note: The bounding box image using red color is the query image.
If there is a bug, don't hesitate to contact me. Thanks!