The demo code to illustrate our paper 'Fusing Global and Local Deep Representation for Effective Object Retrieval'
Our work is based on the work of Tolias of RMAC.
To run the code, you need to download the image of Oxford5k and Paris6k datasets and deep model released by Tolias in their ICLR 2016 paper 'Particular object retrieval with integral max-pooling of CNN activations'.
You need to put the image in datasets directory with sub-direcotory oxford5k and paris6k, where placing images and ground-truth .mat file.
By fusion global and local R-MAC feature, the mAP can be improved and cost no extensive computation.
If you have any problem, please contact me.