Overview

App.py is the entry point. If takes a number of arguments (look at the help function): cd image-detective python DET\app.py -h

Make sure that you always execute the calls in the root directory.

Example usage:

• Generate test images: python DET\app.py generate
• Index all images: python DET\app.py index
• Index selected images: python DET\app.py index -i 2*.jpg
• Query all images with all algorithms using full and feature hashes python DET\app.py query > CSV output in root directory
• Query selected images with all algorithms using feature hashes only python DET\app.py query -i 3*.jpg -ft > CSV output in root directory
• Query all images with the phash algorithm only using full hashes only python DET\app.py query -a phash -fl > CSV output in root directory

Installation (Windows only, linux/mac is much easier)

We need python with the following libraries:

• numpy - Scientific computer python library
• scipy - Scientific computer python library
• pillow - Python imaging library (to create test images)
• imagehash - Python image hashing library
• cv2 - OpenCV python wrapper
• six - Some python utility functions
• elasticsearch - Elasticsearch wrapper
• certifi - Remove SSL exceptions
• requests - http requests
• boto3 - AWS cli

Installing scipy and numpy on windows is a pain. I would strongly suggest installing Anaconda directly instead of python (Anaconda is a bundled version of python that also contains key computer science libraries like numpy and scipy) - https://www.continuum.io/downloads

• Download Anaconda with base of Python 2.7 (although 3.x will probably work, the examples of hashing tend to use 2.7)
• Unzip to a local directory. I use C:\path_installations\Anaconda2
• Add above directory to PATH environment variable
• After all is done, open a cmd and install the above libraries pip install six (You should only need to do pillow, imagehash, six & elasticsearch)

Install elasticsearch - A lucene based search engine similar to solr

• Download from https://www.elastic.co/downloads/elasticsearch
• Follow installation steps (use defaults, add windows service by executing elasticsearch-service install from bin directory)

Install OpenCV - The most popular computer vision library

• Download 2.4.13 from http://opencv.org/downloads.html
• Unzip to a local directory. I use C:\path_installations\opencv-2.4.13
• Copy and paste the file C:\path_installations\opencv-2.4.13\build\python\2.7\x64\cv2.pyd to your python site-packages directory C:\path_installations\Anaconda2\Lib\site-packages

AWS

pip install awscli

Run: aws configure

Access: (ask for this) Secret: (ask for this) Region: eu-central-1

You can now download the source images.

For creating a lambda function, copy the contents from the lib folder into the root and zip the whole lot (remove some of the unneeded folders, reports, images etc). This should run on a lambda function as the python libraries have been pre-compiled and packaged.

IDE

For simple projects like this I tend to you Visual Studio Code as it is basically Sublime Text + git + console + auto-linting + good and quick to install plugins Any decent IDE will work.

Reports

Many as above, although I also tweaked it slightly and there are some older reports in the old_style_reports directory

Limitations

There are many, this is just a starting point...