Weekly update for September 11

Upload your progress (using RStudio) 1/2

• Make sure that you connect RStudio and Github. Click here.
• Go to File -> New Project -> Version Control -> Git.
• Fill out:

• Go to File -> Open file… -> 02_weekly-updates/update-2017-09-11.Rmd

Upload your progress (using RStudio) 2/2

• Add a few slides and save. You can copy the format of other slides or make something more sophisticated.

• Click ‘Knit’ button.

• Commit and push your changes on ‘update-2017-09-11.Rmd’ and ‘update-2017-09-11.html’ in the Git section.

shinyapps repo

The shinyapps repo on github is a sandbox for our different shiny applications:

bulletr, x3p-viewer, and groove-finder are on there at the moment.

ICFIS Workshop

Title: “Elements of Forensic Science for Forensic Practitioners”

Lessons learned:

• Siloization - different disciplines have own language and terms
• “Hierarchy of Propositions”
  • Source level - is a person the source of the evidence? (CSAFE can help most here, imho)
  • Activity level - how did that evidence get there (investigators)
  • Offense level - did a criminal act occur? (judicial system)

ICFIS Workshop, Fingerprints

• Fingerprint evidence protocol ACE-V
  • Analysis - information gathering step (1 + 2 above)
  • Comparison - side-by-side of latent v. known (3 + 4 above)
  • Evaluation - decision making (5 above)
  • Verification - 2nd analyst repeats ACE (6 above)
• GYRO marking of features - subjective qualitative assessment of uncertainty
  • Green = High confidence
  • Yellow = Medium confidence
  • Red = Low confidence
  • Orange = Annotations after viewing known print

ICFIS Workshop, Trace Evidence

Q: What is trace evidence?

A: Microscopic debris easily transferred from one location to another

• “We will attempt to physically match anything”
• Paint, tape, glass, fiber, etc.
• Flowchart of analysis:

ICFIS Workshop, DNA

• PCR-STR: polymerase chain reaction short tandem repeats
• Replicate a DNA sample, count occurrences of STR alleles at 23 loci
• One mismatched loci is enough to eliminate (in single-source sample)
• Markers for sex are DYS391, Chromosomes (X,Y), one other location
• SE33 is a very discriminating location. more info here
• At most 2 values per location - one from mom, one from dad.

MTurk Preliminary Results 1

MTurk Preliminary Results 2

MTurk Preliminary Results 3

Shoeprint 1 by Soyoung

• Created database of 150 pairs of shoe print images with 5 reps on each foot.
• The algorithm speeds up and is packed as the package ‘shoeprintr’.
• Toward a final score for similarity between shoe prints.

Shoeprint 2 by Soyoung

• Find more evidence of KM and KNM.

Jason Saporta

I spent the summer at Credit Suisse making models for predicting loan defaults, prepayments, etc. Not much connection with CSAFE unless you believe everybody on Wall St. is a criminal.

During the past week I implemented a neural network for matching bullet signatures. Everything’s running well, but no serious results yet.

Plans for this week: - Maybe actually getting results for the aforementioned model - Replicating studies on CNN effectiveness for stego problems - Reading more about recent developments in Bayesian ML and determining if that could help with my research

Shoeprints App and Package (Guillermo and Martin)

• Upload your images
• Pre-process image (negate values, rotate, crop, etc)
• Algorithm to find affine transformation parameters

On Shoeprints Profile (Guillermo and Martin)

Regarding Shoeprints Database (Guillermo and Martin)

• A lot of this is going on. Among them
• Matscan (feet pressure heatmap)
• 3D scanner (fixed)
• 3D scanner (handheld)
• High resolution camera
• Shoeprints on sand
• Stand on paper (detailed, walking, stomping, twisting and more than one print)

Amy - Outreach

STEMversity in Milledgeville, GA - 25 high school students
- One day program
- Intro, shoe print activity, handwriting activity

OPPTAG Explorations
- 15 high school students
- One week program
- ``From the Crime Scene to the Court Room’’
- Alicia, Officer Natasha Greene, DCI Crime Lab, Brandon Garrett J.D.

Amy - Outreach

Coming soon
- National Organization for the professional advancement of Black Chemists and Chemical Engineers (NOBChCE)
- Science Bound (September 23)
- Hands on Activity Modules (Handwriting & Shoeprint)

Amy - Handwriting

REU Handwriting Samples - 10 students
- 2 days
- CEDAR Letter & London Letter repeated 3 times each - Signature page

Flash ID

Amy - Handwriting

Automatic Handwriting Matching

• Thinning Image
  
• Vertex Detection

Automatic Handwriting Matching – Code

devtools::install_github("CSAFE-ISU/handwriter")
library(handwriter)
data(london)
london_thin = thinImage(london)
london_nodes = getNodes(london_thin)

Automatic Handwriting Matching – Thinning

plotImageThinned(london, london_thin)

Automatic Handwriting Matching – Vertex Detection

plotNodes(london, london_thin, london_nodes)

Ganesh

Summer: Development of x3p write capability in the bulletr package for different profiles Shiny app (wrApp) for reading dat/x3p files and writing in x3p format - Single file mode - Batch Mode and a bunch of other things

Here’s the link https://github.com/ganeshkiastate/wrApp

Plan for the coming days:

Adaptation of the Chumbley score method from toolmark comparisons to bullets. Key challenge: Understanding the drawbacks of the current algorithm when applied to curved surfaces.