TallyCalculus is an assessment platform by TEPL designed for Tally certification and exam preparation, featuring tools like Question Bank Management and on-screen calculators.
The platform includes proctoring tools to ensure exam integrity, monitoring candidates through webcam feeds, screen activity, and audio analysis.
Challenges like false positives due to environmental noise led to the implementation of smart thresholds and machine learning models in TallyCalculus.
Identity verification was improved with face comparison and liveliness detection to prevent impersonation risks during exams.
A progressive face comparison approach was introduced, capturing fresh images at the start of exams to reduce false positives.
Technologies like TensorFlow and OpenCV were utilized for face comparison and liveliness detection, optimizing the in-house proctoring service.
Privacy concerns were addressed by keeping all exam data within the infrastructure, avoiding external processing in cloud-based solutions.
TallyCalculus stands out for its cost efficiency, data privacy, and customization compared to cloud proctoring services.
While the system excels in identity verification, there are ongoing improvements needed for low light conditions and liveliness detection.
Overall, the new proctoring enhancements have made TallyCalculus more efficient and cost-effective while maintaining exam integrity.