+ - 0:00:00
Notes for current slide
Notes for next slide

Weekly Update for September 17

2018/09/17

1 / 61

About Slides

2 / 61

New CSAFE slide template

We're now using xaringan

What's changed:

  • New person slide: 
---
class: inverse
# Your Name
3 / 61
  • New content slide with title: 
---
class: primary
# Title of slide 
Slide content
  • New content slide without title: 
---
class: secondary
Slide content with no title on slide
4 / 61

Sample User

5 / 61

Sample Slide

  • Sample Table. Sam Tyner talked about the three must-haves of your summary:
Must-have It means
Context Why are you doing what you're doing? "I'm working on X project in order to Y"
Content What are you doing? "I wrote X function that does Y" or "I ran a simulation of Z"
Conclusion What did you learn? "This will help me because it..." or "This important because it gets us to..."
6 / 61
  • If you are going to add an image, create a directory with your name within "images" folder. For example, "images/guillermo/sample_image.JPG"

7 / 61

Amy

8 / 61

Fall Semester Goals

  • Data Collection Application is in 'IRB Staff Review' phase.
  • Expand Bayesian hierarchical model
    • Formal Model Selection Deviance Information Criterion (DIC)
    • Tests and Intervals
    • Validation
  • Write and submit a paper on that modeling.
  • Exploratory analysis of 'next level' features (AAFS in February)
    • Work with Ben and Nick
9 / 61

IRB Manager -
"One form to rule them all"

10 / 61

DIC

Context: Formal model selection with DIC. How many grapheme types to use in modeling?

11 / 61

DIC

Content:

12 / 61

DIC

Conclusion:

  • Work with IT to update server
  • Other model selection techniques.
    • Compositional data.
    • Borrow from topic modeling literature?
13 / 61

Ben

14 / 61

Progress!

  • Checks binary image for predefined patterns.
  • Outputs coordinates for corrections, marked by dots.
15 / 61

  • F. W. M. Stentiford and R. G. Mortimer, "Some new heuristics for thinning binary handprinted characters for OCR," in IEEE Transactions on Systems, Man, and Cybernetics, vol. SMC-13, no. 1, pp. 81-84, Jan.-Feb. 1983. doi: 10.1109/TSMC.1983.6313034
  • Checks for a variety of masks
  • Documentation on my branch (ben)
16 / 61

Nick

17 / 61

Node Count Extraction

Node Counts: 4, 4, 4, 4, 6, 4

18 / 61

Something is strange with the 2 node count -- Need to look at it

19 / 61

Something is strange with the 2 node count -- Need to look at

20 / 61

Nate

21 / 61

Fall Semester Goals

  • Verify theoretical basis for current SLR goals and begin experiments on simulated data
  • Finish groove changepoint detection algorithm version 2 and test on all available data
  • Write paper with Kiegan on groove results
22 / 61

Ganesh

23 / 61

  • Got Comments back from the Reviewers for the first revision, some more minor changes in formatting and language.

    • Fixed most issues and almost ready to send back the responses, pondering on some weird points.

  • Implementing the diagnostics UI (shiny app) on a sample bullet database

    • almost there!
24 / 61

Sam

25 / 61

Book progress

  • Confirmed book contributors:

    • Amanda Luby (human factors)
    • Chris Galbraith (digital)
    • Karen Pan (fingerprint, glass (?))
    • Eric Hare (bullets)
    • Xiao Hui (casings, digital)
    • Nick Berry (handwriting)
    • Soyoung Park (glass, shoes)
    • Peter Vallone, Sarah Riman (DNA)

  • Chapter Outline:

    • Intro. to problem
    • Data collection and cleaning
    • R Packages available & key functions
    • Drawing conclusions (LR?, SLR?, etc.)
    • Case Study

Other: glossary, hierarchy of propositions, LR chapter?

26 / 61

Practitioner survey

27 / 61

Practitioner survey

28 / 61

EDA glass paper

29 / 61

Guillermo

30 / 61

Plans for this Fall

  • Complete adaptation of registration of 3D scans in R and perform analysis

  • Complete shiny app for 2D outsole scans registration

  • Write paper of the database which must include some data analysis

  • Write paper on speaker recognition (w/Vianey Leos)

31 / 61

James

32 / 61

This Fall

  • Steady Progress

    • 27 Pairs left to 3D Scan

  • Will be Using the Gilman Lab for all "four" Prints

  • Mud prints may be taken in Durham due to the 3D Scanner

33 / 61

Kiegan

34 / 61

Fall Semester Goals

  • Submit to AFTE journal - Finishing up edits this week.
  • Complete writing of Chapman & Hall book
    • Finishing Chapter 3 draft this week
  • "Bake-off" of get_grooves methods coming soon.
  • Ran into an issue with robust LOESS on Houston set
    • Reworking functions to iterate how we want.
35 / 61

Locfit.robust updates

36 / 61

Locfit.robust updates

37 / 61

Locfit.robust updates

38 / 61

Danica

39 / 61

Update

  • Submit draft BF vs. LR paper to AOS -> should be done this week!
    • Write a follow-up paper to LPR -> have a nice draft
  • Continue NIJ grant to validate FDE conclusions
    • Started work on the second phase
    • Trying to determine a good way to combine kinematic measures
  • Working on my Top 10 lists
    • Things statisticians should know before working in forensics -> have a nice draft
    • Things forensic practioners should know before working with statistics -> have a rough draft
40 / 61

Miranda

41 / 61

Neural Networks

  • Continue quality control on labeled images
    • Go through annotations to verify all shapes in a given polygon are labeled
  • Implement image augmentation (in particular, stretching) to prevent over-fitting of specific shapes
    • Regular polygons are identified with high accuracy, but elongated polygons are not
    • Very close to being implemented!
42 / 61

Heike

43 / 61

Scans

  • Hamby set 36 scans done, analysis done, results are a bit strange

  • Hamby clone set 224 scans done

  • Hamby set 224 scans done

  • Hamby set 10 arrived from St Louis, in scanning process.

  • several other sets of bullets and cartridge cases

44 / 61

Hamby 224 - Clone

  • Hamby 224 Clone

    • organized as test set: two known bullets from the same barrel are paired with an unknown

    • total of fifteen test sets (one for each unknown): some clones have replicates

45 / 61

Hamby 224 - Clone and Original

Which barrel in the original is Test Set barrel X?

Test set 1 (Clone 224) versus Barrel 1 (Set 224)

Still todo: Matching individual clones and their respective bullets

46 / 61

Scans

Original

Clone

47 / 61

Hamby 224 - Clone and Original

Test set 11 (Clone 224) versus Barrel 1 (Set 224)

Set 224: Bullet 1 - Bullet 2: 0.98

Set/Clone 224: Bullet 1: 0.97

48 / 61

Strange scan

Scan of original and corresponding clone - middle part of clone does not show any striae

49 / 61

Programming

  • bulletxtrctr re-factoring close to done

  • starting on comparisons of before/after feature values

  • x3ptools and bulletxtrctr

50 / 61

Issues!!

51 / 61

Soyoung

52 / 61

Fall semester plans:

  • Shoes
    • Wrap up shoe analysis using edges and SURF
    • Compare the performance with other methods such as phase only correlations, Fourier-Mellin transformation correlation
    • Do analysis with CSAFE shoe data
    • Submit papers regarding shoe analysis soon!

  • Research on "uncertainty pyramid" with Steve and Hari at NIST

    • Glass analysis with Sam
    • Exploratory analysis
    • Various covariance estimations
    • Submit papers!

  • Working on book chapters of glass and shoes funded by ROpenSci Fellowship

53 / 61

Analysis result with CSAFE shoe data

  • All nike shoes with size of 8.5 and 10.5 at time 4 (6 months)
    • KM (716) : Between replicates
    • KNM (599) : Between shoe IDs
    • Edge matching with three circles
54 / 61

Susan

55 / 61

Bullets

  • Hamby 44 analysis
    • Goal is to compare Hamby 36 to Hamby 44
    • Hopefully will be finished later today
56 / 61

Image Alignment

  • ShoeAlignR package renamed to ImageAlignR
    • useful for fingerprints as well
    • algorithms aren't working with color images yet... not sure why
  • Implemented an alignment algorithm from a Kaggle post
    • Harris Corner detection + KNN features + RANSAC to find a matching transformation for alignment
57 / 61

Image Alignment

58 / 61

Image Alignment

  • Orange points - interesting feature for the first image
  • Purple points - similar feature in 2nd image
59 / 61

Image Alignment

  • Red = appears in warped image only
  • Blue = appears in unwarped image only
  • Black = appears in both images (overlapped points)
60 / 61

Image Alignment

  • Still to do:
    • Need to add a better feature detector - corner detection misses some shoe features
    • Figure out whether the algorithm isn't working on certain shoes:
      • Image cleaning?
      • Poor feature detection?
      • Different people wearing the shoes?
61 / 61

About Slides

2 / 61
Paused

Help

Keyboard shortcuts

↑, ←, Pg Up, k Go to previous slide
↓, →, Pg Dn, Space, j Go to next slide
Home Go to first slide
End Go to last slide
Number + Return Go to specific slide
b / m / f Toggle blackout / mirrored / fullscreen mode
c Clone slideshow
p Toggle presenter mode
t Restart the presentation timer
?, h Toggle this help
Esc Back to slideshow