Skip to content

Latest commit

 

History

History
64 lines (30 loc) · 1.34 KB

Fingerprint.md

File metadata and controls

64 lines (30 loc) · 1.34 KB

Step 1: crop fingertip

  1. Get hand key-points with Mediapipe


  1. bring distal phalanges from hand key-points to get fingerprints


  1. extend the fingertip position to the end of the finger.


  1. Get end of the finger and distal phalanges vector to calculate orthogonal vectors, and get coordinate of the box.



Step 2: Label cropped fingertip



Step 3: Train efficientNet for predicting whether fingerprint is exposed or not



Step 4 : Train Auto-Encoder with U-net architecture to manipulate fingerprint



Step 5 : Annotate fingerprint and training segmentation model

After apply those models appropriately, the result is below.



Result

Before


After