Hierarchical Feature For Scene Parsing Using Fully Recurrent Network

In scene parsing, the wide-range contextual information is not effectively encoded. Scene parsing provides segmentation and determines an scene into different regions associated with semantic categories. The main objective of scene parsing is to reduce semantic gap between humans and computer machines on scene understanding. The scenes parsing applications are object detection, text detection on video frames etc. Scene parsing is one of the key problems in computer vision and solved using Recurrent neural network. Labeling small objects is the main difficulties in the scene parsing. The proposed method is Fully Recurrent Network using BiDirectional Recurrent neural network enables the network to model long-range semantic dependencies among image unit. Local representation and Local classification can be enhanced by Directed Acyclic Graph. Training RNN using gradient based method. Back propagation through time algorithm can be used to calculate the gradient method. Recognition accuracy can be improved with the class weighting function.The result can be predicted accuracy using Siftflow data set.