This paper proposes a practical background subtraction approach, NeSe, for tiny energy-harvested sensors leveraging MRAMs. NeSe allows the accuracy and efficiency of event detection to be adjusted at runtime based on the application's requirements. Furthermore, the proposed design reduces data movement overhead due to the near-sensor implementation of background subtraction. Moreover, MRAMs ensure intermittent resiliency, meaning if the power is cut, the background remains unchanged. Finally, if the moving object is detected, the device switches to the high-powered sensor mode.