MAPLES OSM — Digital Answer Script Evaluation Platform

MAPLES OSM is an end-to-end digital answer script scanning and on-screen marking platform built for exam boards, universities, and state education departments. It replaces traditional paper-based evaluation with a secure, scalable digital workflow — from physical booklet receipt through scanning, evaluator assignment, on-screen marking, moderation, and result publication. The platform has processed over 20,00,000 answer scripts across 467 institutions in Tamil Nadu, supported 11,000+ evaluators, and handled 4,000+ concurrent evaluations at peak load. Key capabilities include evaluator anonymity, double valuation with automatic discrepancy detection, face recognition proctoring, multi-layer moderation, and RTI-ready audit trails that log every action with timestamp and user ID.

MAPLES OSM is optimized for desktop browsers including Chrome, Firefox, Edge, and Safari. The evaluation interface requires a minimum screen width to ensure accurate marking with annotation tools, zoom, pan, and question grid navigation. For support staff, a dedicated Android app called MapleSOSM Calls is available — it provides caller ID for 11,000+ evaluator contacts, call queue management, recording upload, and credential resending. The mobile app requires Android 10+ and is only 3.5MB in size. Scanning stations run the MapleScan web app in any modern desktop browser with WebRTC camera support.

Answer sheets are scanned using MapleScan, a purpose-built web-based scanning system. Operators place open booklets under an overhead camera (USB webcam or document camera) on a dark mat surface. MapleScan automatically detects the spine of the booklet, splits the image into left and right pages, applies perspective correction, and validates QR codes and barcodes on each page. Auto-capture mode uses motion detection to photograph pages as the operator flips them — no manual trigger required. Each capture processes in under one second, and a full 32-page booklet can be scanned in under two minutes. All images are stored locally and backed up to Cloudflare R2 in real-time for redundancy.

Yes. MAPLES OSM implements multiple layers of security to protect exam data. Authentication uses JWT tokens with bcrypt password hashing and OTP verification via SMS and email — no shared passwords. Role-based access control with 11 specialized roles ensures users only see data relevant to their function. Assignment-based access means evaluators can only view their specifically assigned answer books. All data is encrypted in transit via HTTPS and stored with dual backup on local infrastructure and Cloudflare R2. Face recognition proctoring continuously verifies evaluator identity during marking sessions. Transaction-isolated database operations prevent race conditions. A complete audit trail logs every action with timestamp and user ID, making the entire system RTI-ready and tamper-evident.

Yes. MAPLES OSM supports double valuation where the same answer book is independently assigned to two or more evaluators who mark without knowledge of each other's scores. Once both evaluations are complete, the system automatically compares scores and flags discrepancies that exceed configurable thresholds. Flagged answer books are routed to moderators for adjudication. Moderators can review both sets of marks side-by-side, verify evaluations, issue warnings for over-valuation or under-valuation, and make final score decisions. This multi-layer approach — evaluators, moderators, chief examiners, and administrators — ensures evaluation quality and eliminates single-evaluator bias.

The MAPLES OSM platform interface currently supports three languages: English, Hindi, and Tamil. The internationalization framework is designed to be extensible, allowing additional languages to be added without code changes to the core platform. Language selection applies to the full evaluation interface including navigation, annotation tool labels, question grids, dashboard headings, and system notifications. This multi-language support is particularly important for state exam boards like DOTE Tamil Nadu where evaluators across 467 institutions are most comfortable working in their regional language.

Quality control in MAPLES OSM operates at every stage of the evaluation pipeline. During scanning, QC operators inspect scanned images in configurable grid views (2, 4, or 8 columns), approving or rejecting pages with structured reasons — rejected pages can be individually re-cropped with live perspective correction without rescanning the entire booklet. During evaluation, moderators review evaluated answer books, verify scores against answer sheets, and flag inconsistencies. Chief examiners oversee entire subjects and monitor evaluator performance. Administrators perform final result approval through a 9-point error validation process before publication. The system automatically issues warnings for statistical anomalies like consistent over-valuation or under-valuation patterns.

Yes. MapleSOSM Calls is a dedicated Android application built specifically for support staff who assist evaluators during exam evaluation windows. The app provides caller ID that instantly displays evaluator names, institutions, and roles from a locally cached directory of 11,000+ contacts. It includes call queue management for organizing support requests by priority, call recording with automatic upload to the central MAPLES OSM platform, and a credential resending feature that lets support staff help evaluators recover login credentials without admin intervention. The app requires Android 10 or later, is only 3.5MB in size, and syncs all data with the web platform in real-time.

Yes, caller identification in MapleSOSM Calls works fully offline using a locally cached contact directory containing 11,000+ evaluator entries including names, institutions, departments, and roles. When an evaluator calls the support helpline, the app immediately displays their identity even without internet connectivity. Call logging, notes, and recording uploads are queued locally and sync automatically when connectivity is restored. This offline-first design is critical during peak evaluation periods when support staff may be in locations with unreliable internet but still need to identify and assist evaluators calling for help.

MAPLES OSM is proven at production scale with over 20,00,000 answer scripts processed, 11,000+ evaluators conducting evaluations simultaneously, and 4,000+ concurrent active evaluation sessions at peak load. The scanning infrastructure supports 50 stations processing 25,000 scripts per day, with load testing verified at 40+ simultaneous stations. During peak evaluation cycles, the platform processes over 10,00,000 pages per day. The architecture achieves this scale through Redis caching for frequently accessed data, PostgreSQL with connection pooling for reliable database access, horizontal scaling with multi-worker uvicorn, and Cloudflare R2 for distributed object storage. Result timelines drop from 90 days to 30 days compared to paper-based evaluation.

Setting up a MapleScan scanning station requires three things: an overhead camera (any USB webcam or document camera with sufficient resolution), a dark mat surface for contrast-based booklet detection, and a computer running the MapleScan web app in any modern browser with WebRTC camera support. No specialized scanner hardware, no proprietary equipment, and no software installation is required — the entire scanning application runs in the browser. The overhead camera approach combined with automatic spine detection means operators simply place open booklets flat on the mat and the system handles page splitting, perspective correction, and QR/barcode validation automatically.

Each capture in MapleScan takes under one second to fully process — this includes spine detection, left-right page splitting, perspective correction, QR/barcode validation, and saving to both local storage and Cloudflare R2 cloud backup. A full 32-page answer booklet can be completely scanned in under two minutes. With auto-capture mode enabled, the system uses motion detection to automatically photograph pages as the operator flips them through idle, turning, settling, and capture states — making the process hands-free. At scale, 50 scanning stations operating simultaneously achieve a combined throughput of 25,000 answer scripts per day.

MAPLES OSM handles poorly scanned pages through its built-in quality control workflow without requiring a full booklet rescan. QC reviewers inspect scanned images in configurable grid views and flag any pages with quality issues such as blurriness, incomplete capture, or incorrect perspective. Flagged pages are routed to operators who can use the manual crop tool — an interactive percentage-based cropping interface with live perspective correction preview. The operator adjusts the crop boundaries, the system applies perspective correction in real-time, and the corrected page replaces only the problematic one. This page-level correction approach saves significant time compared to rescanning entire 32-page booklets.

MAPLES OSM has been load-tested with 40+ concurrent scanning stations all uploading simultaneously to the central server, and the current production deployment supports 50 scanning stations. The system scales horizontally through multi-worker uvicorn for handling concurrent upload requests, PostgreSQL connection pooling for reliable database access under load, and Cloudflare R2 for distributed object storage that handles parallel writes efficiently. Each station operates independently with its own local storage as a first backup, meaning network congestion or server delays at one station never block scanning at another. The architecture is designed to add more stations as needed without configuration changes.

Yes. MapleScan's spine-based detection algorithm works with any open booklet size because it identifies the spine (the center fold) of the booklet rather than relying on fixed page dimensions. The spine offset parameter is adjustable from -30% to +30% to accommodate different binding styles — from center-stapled booklets to side-stitched or perfect-bound answer books. This flexibility means institutions do not need to standardize on a specific booklet format or size to use the scanning system. The perspective correction algorithm also adapts to different camera heights and angles, ensuring consistent output quality regardless of the physical setup at each scanning station.