Using Bayesian neural networks to classify segmented images
Francesco Vivarelli · 1997
We present results that compare the performance of neural networks trained with two Bayesian methods, (i) the Evidence Framework of MacKay (1992) and (ii) a Markov Chain Monte Carlo method due to Neal (1996) on a task of classifying segmented outdoor images. We also investigate the use of the Automatic Relevance Determination method for input feature selection. Using Bayesian neural networks to classify segmented images 2 1 Introduction This work deals with the Bayesian training of neural networks for classifying regions of outdoor scenes. Outdoor scene analysis is usually carried out on images which have been segmented into regions. To carry out the task successfully it is important to classify each region not only using its own attributes (e.g. colour, shape, texture) but also to take account of the context of the other regions. For example, this means that a region surrounded by sky should probably not be classified as a vehicle. Taking account of context can be handled in two way...