User:Saweiner

Project Description

Throughout Marquette University's 10-week summer REU program, Computation Across the Disciplines, I will be researching applications of graph theory to computer vision and image segmentation problems. More specifically, I hope to be able to utilize the graph-cut technique in order to isolate images of the heart and its components from CT scans provided by local hospitals. I will be working under the guidance of Dr. Naveen Bansal.

Week 1 (5/29 - 6/1)

Tuesday 5/29

• Attended REU orientation
• Met briefly with Dr. Bansal to begin planning the summer's research

Wednesday 5/30

• Read introduction to Neural Networks: a Comprehensive Foundation
• Met with Dr. Bansal and graduate student to begin learning about statistical applications to image segmentation

Thursday 5/31

• Met with graduate student to obtain preliminary research publications
• Met with Dr. Bansal to clarify research goals and outline the structure of the project
• Read several research articles and located five articles for further reading

Friday 6/1

• Familiarized myself with several graph-theoretic terms and concepts, in preparation for future research
• Completed Fast Approximate Energy Minimization via Graph Cuts (Boykov et al.)
• Located and obtained the Ford-Fulkerson publication on flow networks in Memorial Library

Week 2 (6/4 - 6/8)

Monday 6/4

• Completed Interactive Graph Cuts for Optimal Boundary & Region Segmentation of Objects in N-D Images (Boykov et al.)
• Completed An Omptimal Graph Theoretic Approach to Data Clustering: Theory and Its Applications to Image Segmentation (Wu et al.)
• Began Graph Cuts and Efficient N-D Image Segmentation (Boykov et al.)

Tuesday 6/5

• Completed Graph Cuts and Efficient N-D Image Segmentation (Boykov et al.)
• Completed Normalized Cuts and Image Segmentation (Shi et al.)

Wednesday 6/6

• Began writing code for GraphCuts algorithm, completed about 350 lines

Thursday 6/7

• Finished a rough draft of the GraphCuts algorithm, totaling about 700 lines
• Attended weekly REU luncheon and learned about the do's and don't's of research presentations
• Met with Dr. Bansal to give a progress report and discuss next steps
• Completed module 1 of the Responsible Conduct of Research training

Friday 6/8

• Completed modules 2 and 3 of the Responsible Conduct of Research training
• Began to modify GraphCuts algorithm (fixing various syntax errors and adding small improvements)

Week 3 (6/11 - 6/15)

Monday 6/11

• Continued to fix GraphCuts algorithm
• Found more efficient storage for Graph data structure
• Fixed several initialization errors
• Added several fringe cases onto the algorithm, now totals about 1200 lines
• Finished second draft of GraphCuts algorithm

Tuesday 6/12

• Met with grad student to discuss generalities of GraphCuts algorithm
• Worked with grad student to implement algorithm

Wednesday 6/13

• Attended the Ethics in Research training session, proctored by Dr. Brylow

Thursday 6/14

• Continued revisions of GraphCuts algorithm in preparation for its first test run
• Met with Dr. Bansal for our weekly progress report
• Took preliminary steps to obtain image data for future use with the GraphCuts algorithm
• Sought out different implementations of the max-flow/min-cut algorithm to compare to (and possibly combine with) my current GraphCuts algorithm

Friday 6/15

• Continued to modify GraphCuts algorithm
• Added condensation method to GraphCuts algorithm in an attempt to conserve memory

Week 4 (6/18 - 6/22)

Monday 6/18

• Read An Experimental Comparison of Min-Cut/Max-Flow Algorithms for Energy Minimization in Vision (Boykov et al.)
• Modified GraphCuts algorithm based on Boykov's active node method of operation
• GraphCuts is use approaching completion, should be ready for testing by next week

Tuesday 6/19

• Fixed several syntax and arithmetical errors in the GraphCuts algorithm
• Met with Dr. Bansal and graduate student to obtain image segmentation data set
• With new image data as reference, I began to write main method for reading in image and implementing the graphCut method (completed about 600 lines)

Wednesday 6/20

• Completed code for handling the graphics of the GraphCuts algorithm, algorithm now totals about 2500 lines
• Encountered import errors involving the graphics package, began to search for an appropriate library download

Thursday 6/21

• Began to create a new Java library to handle graphics of GraphCuts
• Acquired necessary source code for new library from GitHub
• Attended weekly REU luncheon to share progress and witnessed a presentation on computer science education
• Met with Dr. Bansal for our weekly progress report and showed him the GraphCuts algorithm as it currently stands