Fast, Dense Feature SDM on an iPhone 7


Ashton Fagg, Simon Lucey, Sridha Sridharan


In this paper, we present our method for enabling dense SDM to run at over 90 FPS on a mobile device. Our contributions are two-fold. Drawing inspiration from the FFT, we propose a Sparse Compositional Regression (SCR) framework, which enables a significant speed up over classical dense regressors. Second, we propose a binary approximation to SIFT features. Binary Approximated SIFT (BASIFT) features, which are a computationally efficient approximation to SIFT, a commonly used feature with SDM. We demonstrate the performance of our algorithm on an iPhone 7, and show that we achieve similar accuracy to SDM.


  • PDF (ArXiv Version)
  • SCR Code (Matlab/Torch) - This repository contains our prototype code for training and testing our SCR models, however it is not designed to be fast
  • SCR Runtime for iOS - This repository contains our fast implementation of SCR for running on the iPhone. Coming soon.
  • SDM Code (Matlab) - This is our implementation of SDM according to Xiong and De La Torre (2013).
  • Contact

    Ashton Fagg (ashton at fagg dot id dot au)