RMAC image retrieval

RMAC means using RMAC layer as image feature.

Poject structure

➜  rmac_retrieval git:(master) ✗ tree -L 2
.
├── README.md
├── feats
├── model
│   └── pool5_vgg16.prototxt
├── src
│   ├── brute.py
│   ├── extract_features.py
│   └── oxford5k_feats_extract.py
└── tools
    └── compute_ap.cpp

Build evaluation script

The official C++ program provided by the Oxford group, compute_ap.cpp, for computing the mean average precision (mAP) on the retrieval benchmark is provided in this repository for convenience. It is modified to add an explicit include so that it can be compiled everywhere.

g++ -O compute_ap.cpp -o compute_ap

CNN Model

model	Dataset used to fine-tuning	layer	mAP
VGG_ILSVRC_16_layers.caffemodel	-	fc6	45.9%