›CONTENTS/PLAY/DYNETI

Dyneti Work Sample

Dyneti is an AI card scanning company for checkout and fraud prevention.

The problem

Dyneti is an AI card scanning company for checkout and fraud prevention. Its models needed reliable card image data they could train on.

What I did

I worked with a batch of raw card images in .jpg and organized their IDs into the expected file structure. The labeled output I submitted was a set of .txt files, one file per class, each listing the raw .jpg image IDs that belonged in it.

The work sample
01Raw input

The unlabeled batch as it was given: raw card image files.

Raw card images as they were given
02Labeled output

What I submitted: text files matching Dyneti's expected file structure. Each .txt file is a class. I sorted the raw image IDs into good_original.txt for genuine cards, the bad_ files for paper copies, recaptures, and tape, and the unclear files for images classified as unclear.

Raw input - card images

00001916-5fd1-4802-8db5-1fc5b9b7e9f5-0
00027b57-9f61-44fd-b6cb-3ab1e9d1a7b9-0
000379c0-4938-469e-bd7a-df003c0fe059-0
…

Finished product - labeled files

  • good_original.txtGenuine original card photos
  • bad_paper.txtPaper copies of a card
  • bad_other.txtOther invalid or spoofed cards
  • bad_paper_handwritten.txtHandwritten paper copies
  • bad_paper_printed.txtPrinted paper copies
  • bad_recaptured.txtPhotos of a photo or screen
  • bad_physical.txtPhysically altered cards
  • bad_tape.txtCards with tape or coverings
  • unclear.txtToo unclear to classify
  • unclear_recaptured.txtUnclear recaptured images