Skip to content


The coursework is based on gait biometrics. It contains a dataset of people as images from the Southampton Gait Database. These images are from a range of angles.

Need to create a system to recognise subjects by their gait. Have training and test dataset already split for us. Wanted result is to plot inter-, and intra-class based variation histograms and also compute the CCR.

Also plot CMC and ROC curves to calculate EER.

Coursework marking in levels. Simplest is to take manual measurements of people, then use a spreadsheet to do the computations. Using a metric based on the shape of the body. We can take measurements of the body (e.g., length of shoulder, hand, size of head), then use these as features to feed into a classifier. This is the simplest approach, and will get a lower mark.

More sophisticated systems will get a higher mark. For example, can use already published code as a tool to aid development of our own system.

From here, we write a report.

Mark Scheme

  • Report Presentation 25%
  • Recognition Performance 25%
  • Selection and Justification of the Techniques Used 25%
  • Plots, analysis, ROC CMCs given, showing advantages, disadvantages and future directions 25%.

Coursework is over 4 working weeks. Supposed to be 1/4 done before Easter, 3/4 to do after the break.

Word limit 2000 words. IEEE format. Can use existing code or develop own tools, but needs attribution.

Link to spec