bow_image_retrieval

Bag-of-words Image Retrieval

Let's implement traditional way of image search system

Any contributions are welcome :)

Requirements

Tested on Ubuntu 14.04

# OpenCV Contrib
pip install opencv-contrib-python
           
# Scikit for RANSAC
pip install scikit-image

# Datasketch for minHash
pip install datasketch -U

# Config Parser
pip install pyyaml

Modified Datasketch for minHash

You need modified version of datasketch package. https://github.com/insikk/datasketch/tree/update_with_int You can simply install from submodule added in this directory.

git clone --recursive <this repo>
cd datasketch
pip install -e .

Modified pqkmeans

cd pqkmeans
git submodule update --init --recursive # get pybind 
pip install -e .

FLANN and python binding (pyflann)

Download https://github.com/mariusmuja/flann/releases/tag/1.9.1 for python binding. build with cmake

# Install required packages 
sudo apt-get install libhdf5-dev libgtest-dev

# Finish google test install following https://www.eriksmistad.no/getting-started-with-google-test-on-ubuntu/
cd /usr/src/gtest
sudo cmake CMakeLists.txt
sudo make
 
# copy or symlink libgtest.a and libgtest_main.a to your /usr/lib folder
sudo cp *.a /usr/lib

# goto ./lib/flann-1.8.4
mkdir build
cd build
cmake -DBUILD_CUDA_LIB=ON ..
make
sudo make install
# Make sure it builds python binding. 
python
>>> import pyflann
>>> # shows no error

PQk-means

Step by Step TODO

Data preparation

Oxford 5k

Feature Generation

Image Feature Generator Image Feature Descriptor

Visual Words Generation

Generate Visual Words: Vector quantization with several methods

flat k-means clustering: simple, but failed to scale
approximate k-menas
- For nearest negihbor finding, we use randomized kd tree forest. We use FLANN(https://github.com/mariusmuja/flann) implementation

[ ] Enable CUDA support for FLANN during FLANN compilation.

(Optional) MinHash Image Clustering

cd .
export PYTHONPATH=$PYTHONPATH:$(pwd)
python ./src/mhic_seed_generation.py

Search Engine

Approximage nearest neighbor

wwfnwg/bow_image_retrieval