Quick Start

Quick Start

A pass through OMRFlow's whole workflow using synthetic data, so you can see what it does without needing real answer sheets. Allow about 30–45 minutes the first time.

Everything here uses data OMRFlow generates itself. Nothing in this guide requires, or should be done with, real candidate information.

Prerequisite: OMRFlow installed and launched.

---

1. Create a project

A project is one folder holding one examination: its database, its imported scans, and its generated workbooks.

1. On the Project stage, click Create Project.
2. Choose a parent folder and a project name, e.g. Demo Exam.
3. OMRFlow creates the folder and opens the project. The title bar and the status bar now name it.

Project Configuration opens automatically for a new project.

2. Describe the examination and its sets

In Project Configuration (also at Application menu → File → Project Configuration…):

1. Set the examination name — the title that appears on reports. It need not be a legal folder name.
2. Define the question-paper sets. If your examination has one paper, one set is enough; give it a code such as A. If candidates sat different papers, add one set per paper.
3. Save.

Sets matter later: attendance is per set, answer keys are per set, and each set's report is built from that set's own attendance workbook. See Examination Sets.

3. Open a template

A template tells OMRFlow where the bubbles are on the page. OMRFlow ships an example, which is the quickest route through this guide.

1. Go to the Template stage.

2. Click the open button on the toolbar (second icon) and choose examples/templates/100_question_4_choice_example.omrt from the repository, if you have it.

   Installed from the installer and have no repository checkout? Download that file from the repository and open it.

3. The canvas shows the reference sheet with its regions outlined, and the Regions list on the left names them: registration markers, an orientation mark, Student ID, Question set, and four question blocks.

4. Click Validate to confirm the template is internally consistent.

To build a template from your own blank sheet instead, see Template Design — that is the real starting point for a real examination, and it is the stage that takes the most care.

4. Generate synthetic answer sheets

Rather than scanning paper, have OMRFlow render sheets it already knows the answers to.

1. Application menu → Tools → Developer / Testing → Generate Synthetic Test Dataset…
2. Choose the same template, an output folder, and a modest count — 25 sheets is plenty.
3. Generate. OMRFlow writes the images plus a manifest recording exactly what it drew on each one.

This is also the right way to produce an attachment for a bug report: the sheets are realistic and contain nobody's data.

5. Process the sheets

1. Go to the Scan stage.
2. Load Template… and select the same template. The panel names it and warns that it has not been calibrated against representative scans — expected here, because synthetic sheets are rendered from the template itself.
3. Add Folder… and choose the folder you just generated.
4. Click Process All.

The progress panel shows completed and total counts, elapsed time, an estimate of the time remaining, throughput, and per-outcome tallies. The table fills in as sheets finish: original file, roll number, set, status and output filename.

Sheets OMRFlow could not decide confidently are marked for review rather than guessed at. Click any row to see the corrected sheet and the recognised values beside it.

For what the status values and answer symbols mean, see Recognition Symbols.

6. Resolve what recognition was unsure about

1. Go to the Resolve stage.

If the queue is empty, recognition was confident about everything — likely with clean synthetic sheets. Skip to step 7.

Otherwise, for each conflict:

2. The queue lists what needs a decision. What the machine saw shows the original sheet, the corrected sheet and a zoomed view of the field in question.
3. Under Your decision, either Accept machine value, enter the right value and Save value, or Defer it.
4. A reason must be chosen, and your reviewer name must be set (Application menu → File → Settings) before a correction can be saved — every correction is recorded against a named person.

Nothing is overwritten. The machine's original reading is kept alongside your decision, append-only. See Review & Resolution.

7. Attendance

Attendance is per set, and the workbook you import becomes that set's result template — so its layout is what your final report looks like.

1. Go to the Attendance stage and pick a set under Attendance by set.
2. Click Download Sample Template… to get a correctly shaped workbook, and open it.
3. Fill in a few fabricated candidates whose roll numbers match the synthetic sheets you generated. (The Scan stage's table shows the roll numbers that were recognised.) Mark one candidate absent.
4. Choose / Replace Attendance File… and select your workbook.
5. Click Reconcile.

The Reconciliation summary reports matched scripts and exceptions: unknown candidate IDs, duplicates, candidates with no script, and absentees that nevertheless have one. Resolve each under Your decision — Assign Script, Set Script Aside, Override Attendance or Accept As-Is.

See Attendance and Attendance Workbook Format.

8. Answer key

1. Go to the Answer Key stage and select the set.
2. Enter the answers. For the example template that is 100 questions; the synthetic dataset's manifest records what was drawn, so you can make a key that produces sensible marks. Alternatively use Read From Solution Sheet… to recognise a key from a filled-in sheet.
3. Save As New Revision.
4. Verify Answer Key. A key must be verified before results can be calculated — a wrong key silently mis-marks a whole cohort, so this is a deliberate second pair of eyes.

Under Wrong questions (full credit for everyone) you can mark a question as defective; every candidate then receives credit for it.

9. Score

1. Go to the Results stage.
2. Scoring Configuration…: marks for a correct answer, for an incorrect one, and for a blank; and whether negative marking applies.
3. Check Before Scoring reports what is not ready — an unverified key, an unreconciled roster, unresolved conflicts.
4. Calculate Results.

The Results summary lists candidates with their marks and ranks. Select a candidate to see their per-question detail; Review Answers… shows it against the key. Recalculate This Candidate re-scores one person after a correction.

Ranking uses standard competition ranking, so 90, 88, 88, 85 gives ranks 1, 2, 2, 4. See Answer Keys & Scoring.

10. Generate reports

1. Go to the Reports stage.
2. The set's attendance workbook is normally already bound as its result template. Select Template… if you need a different one.
3. Validate checks the template against the roster and reports what is missing.
4. Preview shows what will be written.
5. Generate XLSX writes the workbook — a roll-wise sheet built on your own template, a merit-wise sheet with absentees removed and the rest ordered by result, plus Summary, Answer Key and Processing Log sheets.
6. Generate All Sets does every set at once.
7. Generate PDF additionally needs LibreOffice installed; without it the XLSX is still produced.

Your workbook's formatting, fonts and logo are preserved — the report is built by copying your template and writing into the copy. See Results & Reports.

11. Find the logs and support information

• Application menu → Tools → Create Diagnostic Bundle… collects the version, platform, settings, a project health check and the log tail into one zip. It contains no candidate data by design, and is the right thing to attach to a bug report.
• Logs: %LOCALAPPDATA%\OMRFlow\logs\ for the application, and logs\ inside the project folder for that project.
• Application menu → Tools → Project Health / Recovery… checks database integrity and manages backups.

---

What to do next

• Read Known Limitations before trusting any of this with a real examination.
• Build a template from your own blank sheet: Template Design.
• Understand what the Calibrate stage is for before you run a real batch: Processing. Synthetic sheets do not need it; real scans do.
• If something went wrong: Troubleshooting.