Header image  
a web-application for marine biologists  
 
 
Welcome to the PAS homepage!

PAS is a web-application being developed to address the challenging and dynamic problem of automatic classification of plankton images. It is intended to be used by marine biologists to classify their images. It provides the functionality for experts to upload their images, process them, extract features, hand-label exemplars, train classifiers and use those classifiers to automatically label new images. It also has a suite of visualization and analysis tools to aid experts in understanding the errors their classifiers are making to further improve their results.

Our simple software design philosophy was to provide the expert a suite of flexible tools instead of a black box solution. At its core, PAS manages a filesystem and provides a web-interface for the user to manipulate it. A filesystem is our notion of a database and it stores, the images (and their meta-data), the algorithms, the output of any algorithm run on the images, and experimental results.

It is expected that PAS will be made available for download in May 2009. More information, tutorial videos and demos of PAS will be made available on this page upon release.

Key characteristics:
  • Implemented as a web-application that can run on a web server or locally
  • Built around existing popular software packages including ImageJ and Weka
  • Extensive classifier error analysis tools
  • Open source (GPL) and cross-platform (Windows / Mac / Linux)
Main functionalities:
  • Uploading collections of images and their meta-data
  • Filtering data sets based on meta-data
  • Adding algorithms to ImageJ or Weka to be used within PAS
  • Running algorithms on image data sets
  • Importing and exporting images and features
  • Performing classification experiments
  • Data set viewer to be able to search and browse through data sets
Who can benefit from PAS

Simply anyone interested in classifying images. While PAS was developed with marine biologists in mind, there is nothing about this application that restricts it to plankton data. Experts in any other domain interested in classifying image data can benefit from using it. Our core goal was to develop a platform where algorithms and research findings in computer vision and machine learning can be made available to scientists in other areas such as marine biology and systems biology.

Credits

PAS is the software outcome of a collaborative plankton classification project between: This project is funded by the National Science Foundation under grant ATM 0325167.

It is being developed by Marwan Mattar with the help and support of several project members. Please send him email () if you have any inquiries or would like to be added to the PAS mailing list.

PAS relies on the following great open-source software packages:



last modified on Februrary 22, 2009