CNNImageRetrieval: Training and evaluating CNNs for Image Retrieval

CNNImageRetrieval is a MATLAB toolbox that implements the training and testing of the approach described in our ECCV 2016 paper:

CNN Image Retrieval Learns from BoW: Unsupervised Fine-Tuning with Hard Examples, Radenović F., Tolias G., Chum O., ECCV 2016 [arXiv]

What is it?

This code implements:

1. Training (fine-tuning) CNN for image retrieval
2. Learning supervised whitening for CNN image representations
3. Testing CNN image retrieval on Oxford5k and Paris6k datasets

Prerequisites

In order to run this toolbox you will need:

1. MATLAB (tested with MATLAB R2017a on Debian 8.1)
2. MatConvNet MATLAB toolbox version 1.0-beta24
3. All the rest (data + networks) is automatically downloaded with our scripts

Execution

Run the following script in MATLAB:

>> run [MATCONVNET_ROOT]/matlab/vl_setupnn;
>> run [CNNIMAGERETRIEVAL_ROOT]/setup_cnnimageretrieval;
>> train_cnnimageretrieval;
>> test_cnnimageretrieval;

Citation

If you use this work please cite our ECCV 2016 publication:

@inproceedings{RTC16,
 title = {{CNN} Image Retrieval Learns from {BoW}: Unsupervised Fine-Tuning with Hard Examples},
 author = {Radenovi{\'c}, F. and Tolias, G. and Chum, O.},
 booktitle = {ECCV},
 year = {2016}
}