Project P13:
Aligning Images of Flies

Project Goal

The aim of this project is to align images of insects. At Oregon State University's Insect Identification for Environmental Monitoring and Ecological Science Web site you can find images of specific insects (click on 2004-01, and then Cal??). All of the CAL specimens are from the Genus Calineuria, which is a kind of "stone fly". You are asked to take images of the Genus Calineuria fly, and write a computer vision program for aligning them reliably. Here are two examples (click on the images to enlarge):

As a starting point for the type alignment techniques that should be applied, check out this recent paper.

This project will be carried out in collaboration with researchers from Oregon State; point of contact there will be Prof. Tom Dietterich.

Project Scope

To develop software for aligning insects of a single species in photographs. This entails getting the images, and implementing a simple template search for finding parts of the insect. Then implement the correlated correspondence technique for resolving the global alignment of the insect, providing for possible occlusion of insect parts. The outcome should be a set of parameters that specifies how a generic image of the insect has to be morphed to generate the configuration within a specific image.

Tasks

• Download the data, handlabel parts, identify a reference image
• develop template matcher for finding local parts (or maybe use SIFT features)
• develop probabilities for pairs of adjacent parts
• implement inference technique for aligning those pairs with a reference image of the insect

Project Status

Steve Chuang (stevechuang at gmail dot com),
Jessica Yuen (jyuen at stanford),
James Balfour (jbalfour at stanford),
Catherine Chang (catie at stanford)

Point of Contact

Sebastian Thrun

Midterm Report

not yet submited

Final Report

not yet submitted