I think manual attendance registers are outdated and prone to 'proxy' marking by students. To automate this process, we designed a smart attendance system that handles face recognition and liveness detection in real time. By implementing a majority-vote trick across several video frames, we try to eliminate the 'flickering' identity issues common in older software. The project tries to build to run on a standard CPU without needing expensive GPUs, and it still achieves a smooth 18 FPS. These results, including an 84.1% accuracy for masks, which is best for a college project, demonstrate that the tool is a viable, tamper-proof alternative for university environments like Galgotias.
Mehra et al. (Thu,) studied this question.