ClassCloud: A Cloud-Based Optical Mark Recognition System for Periodical Reporting at Baliwag North Central School

Description

A teacher stands at the front of the class with the answer key. “Item one, who got it right?” Hands go up, they count them, move to item two, repeated for every item, every exam, every section, every quarter. The reports that come from this don't stay in the classroom. DepEd uses them to shape national education policy, so a miscount at one desk can distort a decision made far above it.

Some teachers at Baliwag North Central School tried to shortcut this with ZipGrade, a paid scanning app. It tallies correct answers, but doesn't consolidate reports across sections and grade levels, and its features are subscription-gated, so results still get copied into Excel by hand. The manual step it was meant to remove stays exactly where it was.

I led a four-person team as Project Manager and Lead Developer, running Scrum through Jira. Browser-based OMR scanning was unproven when we started, so I set a six-week cap to test it before committing the rest of the timeline to it. We had it working inside that window.

As Lead Developer, I designed the database schema and a Role-based Access Control (RBAC) system enforced through JWTs and Supabase Row-Level Security (RLS), with the audit logging and consent capture RA 10173 requires. The OMR pipeline runs entirely client-side, using OpenCV.js in a browser Web Worker, so the image never leaves the device; only the extracted answers do. One atomic PostgreSQL RPC generates the Item Analysis, Level of Proficiency, and LAEMPL reports on exam finalization.

We tested with teachers after every major sprint. School personnel scored the finished system 3.87/4.0 on the Technology Acceptance Model; IT professionals rated it “Excellent” on seven of eight ISO/IEC 25010 characteristics.

Nobody has to count raised hands to build a national report anymore. The scan does it in the time it takes to say “item one.”