top of page
IMG_2806.JPG

I work with an amazing team at Apple Inc as a Machine Learning Manager. My team is focused on an intersection between state of the art research and products. I received my PhD in computer science in 2016 under Dr. Mubarak Shah at the Center for Research in Computer Vision. 

Latest News

Sep 2018: I started managing a team at Apple working on cutting edge AI.
August 2017: I joined Apple Inc as a computer vision scientist. 
July 2017: Check out our crowd tracking dataset [Link]
April 2017: A story on our crowd tracking work in Science Magazine (AAAS) [Link]
March 2017: Won the 2016-17 university award for Outstanding Dissertation in the category Engineering, Physical Sciences and Life Sciences [Link]
March 2017: Crowd tracking paper is accepted for publication in IEEE PAMI.[PDF]
March 2017: Check out our new license plate recognition software. [Paper on arXiv ][Try the Demo]
Feb 2017: Sighthound releases state of the art age, gender and emotion recognition. [Paper on arXiv ][Try the Demo]
Dec 2016: Sighthound launches vehicle recognition in the cloud [Paper on arXiv]Try the Demo]

Research Highlights

DAGER: Deep Age, Gender and Emotion Recognition

State of the art age, gender and  emotion recognition. Outperforming competitors including Microsoft by large margin. 

Crowd Tracking

One of the very few available trackers that can track hundreds of people simultaneously.

GMMCP Tracker

Data association in this work is formulated as a Generalized Maximum Multi Clique problem (GMMCP). This is an improved version of our previous work, "GMCP" tracker.

Car Make/Model Rec

Car Make/Model Rec

State of the art both in commercial software and academic papers. Checkout paper for results on compCar and Stanford datasets.

Deep Face Recognition on iOS

At Sighthound Inc we have developed a state of the art facial recognition algorithm that runs real time on your iPhone. The algorithm consists of three components, detection, tracking and recognition. All are designed and implemented in house.  

Who's Your Daddy?

TINF Tracker

In this work we consider data association and detection as one single problem. Our framework combines multi-commodity network flow with structure learning to address the multi-target tracking problem.

Who's Your Daddy?

In this study we answer four key questions:

-Do offspring resemble their parents?
-Do offspring resemble one parent more than the other?
-What parts of the face are more genetic?
-Do anthropologies' studies help learn better features?

Visual Business Recognition

A fully automated method for recognizing businesses using visual data along with GPS coordinates of the camera. 

My PhD Thesis

bottom of page