NOvA is a long-baseline neutrino oscillation experiment that uses Near and Far detectors to measure electron neutrino appearance and muon neutrino disappearance. The classification of final sate particles in the neutrino interaction helps us to determine the neutrino flavour. So, NOvA has developed a Convolutional Neural Network for single particle classification which employs context-enhanced inputs. The first implementation of this network was trained on neutrino and antineutrino datasets separately. In this work, we train the network on a combined neutrino and antineutrino dataset and compare with the separately-trained networks. The results show the combined network performing comparatively to the separate networks with chance of improvement with more data. In this work, I will show the comparison of these networks and their performances.