Computerized Image Labeling using Prominent Features of Unidentified Image

Vigorous learning is useful in situations where labelled data is scarce, unlabelled data is available and labelling has some cost associated with it. In such situations energetic learning helps by identifying a minimal set of items to label that will allow the training of an effective classifier. Most existing image processing applications are designed for small-scale and local computation which does not scale well to web-sized problems with their large requirements for computational resources and storage, since a large number of images are not labelled properly which is considered as weakly labelled and some of the images were not labelled at all. In this Map Reducing concept is used to overcome this massive labelled and un-labelled analysis problem, since traditional data processing application software is inadequate to deal with them. In this a novel method is developed for achieving multi-label, multi-feature image annotation using a Map Reduce framework, where an image-level labels and region-level labels for both labelled and unlabelled images are obtained. The associations between semantic concepts and visual features are mined both at the image level and at the region level through which Multi-Label image correlations are obtained by a co-occurrence matrix of concept pairs using Convolution Neural Network.