turkmarker

Image landmarking system for Amazon Mechanical Turk


Keywords
image, landmarking, Mechanical Turk, amazon, amazon-web-services, facial, javascript, mechanical, python, python3, turk
License
BSD-3-Clause
Install
pip install turkmarker==1.0.1.dev1

Documentation

MechTurkLandmarker travisCI codecov

Manual Image Landmarking in Amazon Mechanical Turk

MechTurkLandmarker is a web-based image landmarking tool intended for use with Amazon Mechanical Turk. With MechTurkLandmarker you can easily submit tasks to Amazon Mechanical Turk, allowing for many people to manually identify landmarks of interest on an image. This automatically 'package' generates all of the required files and automates the build and deployment process through through pre-configured Python scripts. The default setup provides all of the files required in order to create a Mechanical Turk HIT to landmark an example face with the MULTI-PIE landmark configuration. An example of a deployed HIT to Amazon S3 can be found here

Installation

Currently the source has not been organised into a single Python package for download from pypi, this may occur in the future with continued development, if there is sufficient demand. For now, simply clone the entire source directory and install all the requirements. It is always recommended to use a virtualenv when installing all requirements.

git clone git@github.com:benjohnston24/MechTurkLandmarker.git
cd MechTurkLandmarker
pip install -r requirements.txt

Configuration Files

  • .configrc - configuration file for the system, modify this file to change aspects of operation and details of the Mechanical Turk HIT tasks or S3 Buckets.

Building

TODO

  • Improve test coverage - Python
  • Add javascript unit tests
  • Add multi-image HITS