batzner / patrec

Flask Web App for pattern matching

PatRec - Pattern Recognition for Logo Matching

PatRec is a Flask Web App that allows you to scan a set of images for a given pattern.

It uses an ensemble of Feature Descriptors to identify the pattern in each of the given images. Thus it achieves a high accuracy without having to train the ensemble.

Installation

There are three options to run PatRec in your browser:

1. Docker (Docker Hub) - (recommended)

Run docker run -p 5000:80 jorba/patrec. This will automatically pull the latest patrec image from Docker Hub (http://hub.docker.com). The image size is 3 GB. This option is the fastest way if you do not have OpenCV 2.4.X installed.

2. Docker (local)

1. Build the image by running docker build -t patrec . (Don't forget the point at the end of the command) in the project's root directory where the Dockerfile is located. However, building the Dockerimage will take a very long time.
2. Run the image with docker run -p 5000:80 patrec

3a. Install without Docker (Linux)

1. Install all python dependencies by running sudo pip install -r requirements.txt in the project's root directory.
2. Install OpenCV 2.4.13 on your machine following these steps.
3. Start the Flask web app with python runlocal.py.

3b. Install without Docker (Windows)

1. (Recommended) Install the Anaconda Python Distribution
2. Install all python dependencies by running pip install -r requirements.txt in the project's root directory.
3. Install OpenCV 2.4.13 on your machine by downloading the OpenCV Installer from the OpenCV Website.
4. Navigate to the folder <Your_OpenCV_Path>/build/python/2.7 and copy the file cv2.pyd to your python 2.7 site-packages folder.
5. To check if the installation was successful run python, import cv2 and cv2.__version__. The version should be 2.4.13.
6. Finally, start the Flask web app with python runlocal.py from the application folder.

Usage

To try out PatRec, go to localhost:5000 and drop the images and the pattern in the boxes. Then click Find matches to see the result.

The result will contain the number of votes from the Feature Descriptors and the corresponding prediction: Match or No Match.