Code

We encourage open source development, and as such our most recent work can be found on Bitbucket. Other work can be downloaded below.

  1. Spatio-Chromatic Opponent Descriptors, This code provides the dense colour-opponent descriptor sampling as described in our 2014 ECCV paper. The latest code is available from this repository. Non-Git users can download the files here., Oct-2014, 


  2. Multiscale Medial Response, “This code demonstrates the use of gradient fields to steer a spatial filtering process.  It is highly unusual in that the calculation of medial response is achieved by synthesizing a steered kernel according to boundary direction.  A unique demonstration of steering, it is also a very elegant computational architecture.”, Nov-2013, 


  3. Complex Gradient Field, “Code for generating and displaying a complex gradient field. This is basic stuff: the complex representation is just a way of encoding magnitude and direction conveniently in a language or architecture supporting complex-arithmetic, such as Matlab. It can also be used for complex template matching with vector fields i.e. incorporating direction directly into correlation-based template matching . The gradient field can also be used to steer subsequent processing (see #2) and, of course, forms the basis  for HoG-based descriptors.” , May-2012, 


  4. Generalised Test Images, “Code for generating synthetic images. The code  is used for creating simple test image shapes that have controllable degrees of spatial blur. This can be used for generating data that is spatially regularized in the sense of being differentiable, and can be tied to analytic models of spatial intensity patterns. I have also found these useful for building spatial kernels, or simple shapes that can be used for hypothesis testing. Examples of usage are provided.”, Apr-2012,