Facial attributes recognition, Cellebrite

Tal Toledano

Data Science Fellows June 2020 Cohort


With increasing hard drive capacity and faster bandwidth, average users are able to correspond and accumulate large date types in the form of images and videos. In order to accelerate investigation to go over large amounts of these specific data types on a personal digital data, confiscate from a suspect in a crime scene. Enters machine and deep learning techniques we utilize for the project in order to productize a facial attribute classifier.

Challenges (at least two)

  1. Handling big data
  2. Serve the model beyond rest api 
  3. working with aws

Achievements (according to KPIs)

  1. Fully achieved project from scraping data to product.
  2. Reach some high performing models 
  3. Built a fast and comprehensive code library with a coherent pipeline.

Future project development 

Adding tools for data generation with focus on specific facial attributes i.e. gang face tattoos.

Improving certain models’ errors and inference run time.

Share this post

Share on facebook
Share on twitter
Share on linkedin
Share on email