We present a novel approach using a fully convolutional neural network (FCNN) for prostate segmentation. The FCNN model used here is a modified version of V-net with smart factorization of convolutional layers for reducing the number of parameters, and Batch Normalization for better and faster convergence. Unlike previous approaches which focus on segmentation using 2D slices, whole volume in 3D has been considered which saves time during inference and additionally, the algorithm can take advantage of the 3D spatial information. Various on-the-fly augmentation methods have been applied to prevent overfitting of the network. The model achieved a Dice Similarity Coefficient (DSC) of 0.92 and 0.87 on training and validation set respectively.