First-Person Activity Forecasting with Online Inverse Reinforcement Learning | Read Paper on Bytez