$\pi^{0}$ mass reconstruction in NOvA Far Detector.

NOvA is a long-baseline neutrino oscillation experiment with functionally identical, segmented, tracking calorimeter Near and Far detectors. The detectors lie 14.6 mrad off-axis from the Fermilab NuMI beam, with a well-defined peak in neutrino energy at 2 GeV. The absolute calibration of the energy scale of the detectors is a major systematic uncertainty in long-baseline oscillation search in NOvA. Neutrino detectors make use of some standard candles for absolute energy calibration. Stopping muon energy distributions, Michel electron energy distributions, and invariant $\pi^{0}$ mass are among them. In this talk, we cover NOvA$^{'}$s use of a new method to identify $\pi^{0}$ with cosmic origins in the NOvA Far Detector. We employ a computer vision based particle identifier using convolutional neural networks (CVN) to identify $\pi^{0}$s, complementing an existing strategy to identify $\pi^{0}$ from the neutrino beam using more traditional methods in the Near Detector.