Image annotation
Bounding boxes, polygons and keypoints for project-defined object classes. Agree label definitions, edge cases and required export format before work begins.
FOR AI DEVELOPERS & TEAMS
Data annotation support.
A clear path from sample to delivery.
IA Data Solutions (SMC-Private) Limited is a Pakistan-registered company developing practical annotation and data preparation services. Start with a scoped paid pilot to assess fit before a larger commitment.
CAPABILITIES
Support matched
to your dataset.
Scope, staffing and turnaround are confirmed after reviewing your sample and instructions. Label Studio is used in our annotation practice.
Document & invoice OCR
Labelling document regions and extracting agreed fields such as invoice numbers, dates and totals, with field rules and ambiguous values reviewed during the pilot.
Video annotation
Object tracking with attention to object identity, frame coverage and occlusion rules. Frame rate, interpolation and output expectations are agreed per project.
Audio & structured data
Speaker segmentation and labelling, plus Excel or Google Sheets support for organising and reviewing data. Language coverage and specialist requirements are confirmed separately.
DEMONSTRATION WORK
Review a sample.
Then define your pilot.
Practice examples include image annotation, invoice OCR, video tracking and audio speaker segmentation.
Request relevant examples
Tell us your annotation type and preferred output. We can discuss which demonstration screenshots and available exports are relevant to your use case.
Request sample work ↗A pilot with agreed acceptance
Agree a limited batch, guidelines, review criteria, price and timeline in writing. Review the result together before deciding whether to expand the work.
Discuss a paid pilot ↗PRACTICE SAMPLE · REVIEW DRAFT
Vehicle polygon annotation portfolio
Compare the original scene with six selected vehicle annotations. The presentation includes individual vehicle close-ups and explains contour tracing, occlusion handling, visible tyre coverage and shadow exclusion.
This is a practice demonstration under review. It covers selected vehicles, with class and roof-rack details still to confirm. It is not a completed client project or a claim of measured accuracy.
VIDEO ANNOTATION · PORTFOLIO DEMONSTRATION
Follow the movement.
Inspect every track.
Explore five vehicles tracked through an aerial road scene, with coloured bounding boxes and consistent vehicle identifiers. This revised visual demonstration presents Full HD source footage with IA Data Solutions branding and a clear playback timeline.
Vehicle 01Vehicle 02Vehicle 03Vehicle 04Vehicle 05
Skills demonstrated
- Separate object identities across a moving scene.
- Bounding-box placement across the vehicle movement.
- Partial visibility at entry and exit boundaries.
- Colour-coded tracks and numbered vehicle identifiers.
- Annotation review using source footage and a Label Studio JSON reference.
Review criteria for a production pilot
- Tight boxes around the visible vehicle body.
- Frame-by-frame checks for drift and missed coverage.
- Consistent class definitions and object identifiers.
- Agreed occlusion, truncation and shadow rules.
- Frame-rate alignment and export validation.
About this sample: A portfolio demonstration based on aerial footage and a Label Studio annotation reference, with revised box placement and timing. Vehicle 01–05 are track identifiers within the Vehicle label. Project-specific classes, export formats and quality criteria are agreed during a pilot. This sample is not a completed client project or a measured accuracy result. No audio is included.
DOCUMENT ANNOTATION · PDF & INVOICE OCR
From invoice text to structured fields.
A fictional invoice sample demonstrating field-level document annotation in Label Studio. Compare the document with its labelled regions and matching text values.
| Label | Transcribed value |
|---|---|
| Company Name | Bisha Office Supplies |
| Invoice Number | INV-2026-001 |
| Date | 06 October 2026 |
| Total Amount | 402.50 SAR |
Skills demonstrated
- Locating key fields within an invoice layout.
- Assigning field-specific labels to text regions.
- Transcribing values from the source document.
- Visual comparison of source text and annotations.
Project requirements we can discuss
- Required fields, document types and languages.
- Date, currency and number formatting rules.
- Handling missing, unclear or repeated values.
- Export schema and review criteria for a pilot.
Fictional practice document. This portfolio shows screenshots of document annotation, not an automated PDF extraction system or a measured OCR accuracy result. Click either image to inspect the full-size screenshot.
AUDIO ANNOTATION · SPEECH SEGMENTATION
Make speech boundaries visible.
A Label Studio practice example showing a speech waveform divided into four labelled regions. The displayed regions all use the label “Speaker one,” with pauses visible between selected speech intervals.
Skills demonstrated
- Reviewing speech activity on a waveform timeline.
- Defining separate time-bounded speech regions.
- Applying a consistent speaker label.
- Inspecting speech intervals and intervening pauses.
Project requirements we can discuss
- Speaker-label definitions and segment boundaries.
- Rules for silence, background noise and overlap.
- Whether verbatim transcription is required.
- Timestamp precision, export format and review criteria.
Screenshot demonstration of speech segmentation. The original audio and a transcript are not included. The available “Speaker two” label is not used in the four displayed regions; this sample does not demonstrate separation of two actual speakers.
KEYPOINT ANNOTATION · VEHICLE & HUMAN POSE
Landmarks that give images structure.
Explore two separate practice examples: vehicle landmarks and human pose keypoints. Each uses a defined label set and manually placed points in Label Studio.
Wheel and headlight centres, windscreen corners and visible mirror centres.
Nose and paired shoulders, elbows, wrists, hips, knees and ankles, using the subject's left and right.
Skills demonstrated
- Landmark localisation and manual point placement.
- Consistent category assignment.
- Visible-feature selection and visual review.
- Subject-based left/right labels for human pose.
Requirements for a pilot
- Agreed landmark definitions and point ordering.
- Rules for hidden or ambiguous landmarks.
- Required output schema and coordinate format.
- Review criteria and revision scope.
Practice demonstrations, not completed client projects or measured accuracy results. Vehicle source image is AI-generated. Human joint positions are visually estimated through clothing.
OBJECT DETECTION · BOUNDING BOXES
Separate objects. Clear boundaries.
Two street-scene practice examples labelled in Label Studio. The first shows three selected vehicles; the second shows a foreground white sedan and smaller background vehicles. The boxes demonstrate individual object localisation, overlapping rectangles and handling of a vehicle truncated by the image edge.
Practice example showing vehicles at different scales with partial occlusion. Four Car regions are listed in the supplied screenshot; individual region validity remains to be confirmed during export review.
Skills demonstrated
- Creating a separate bounding box for each selected object.
- Allowing box overlap where vehicle extents overlap.
- Reviewing visible-object coverage and box boundaries.
- Clipping a truncated object to the image boundary.
Requirements for a pilot
- Vehicle class definitions, including cars and vans.
- Visible-only versus full-object bounding-box rules.
- Minimum object size and occlusion requirements.
- Completeness checks against an agreed scope.
Selected-object practice sample, not exhaustive annotation of every object in the scene. The motorcycle is outside this sample's labelled scope. The displayed vehicle regions use the existing Car label; production class definitions would be agreed before a pilot. No measured accuracy or completed client engagement is claimed.
These are demonstration and practice examples, not completed client case studies. No client results, accuracy percentages or delivery capacity are implied.
PROPOSED PROJECT PROCESS
Agree the details.
Check the result.
Each project starts with written requirements and a manageable initial scope.
Understand the data
Review sample data, label definitions, volume, tools, export format, timeline and access restrictions.
Calibrate on a batch
Agree paid pilot terms and resolve unclear examples before expanding production.
Review against guidelines
Apply the agreed checks, log issues and correct deviations within the agreed revision scope.
Validate the handover
Check the export format and completeness, deliver through an agreed channel and collect acceptance feedback.
QUALITY & DATA HANDLING
Define expectations
before sharing data.
The following is our proposed project framework. Specific controls, review coverage and responsibilities must be agreed before access is granted.
Quality planning
- Document label definitions, exclusions and difficult examples.
- Agree sample acceptance criteria and review coverage.
- Check completeness, label consistency and annotation boundaries.
- Track questions, guideline changes and corrections.
- Agree how quality will be measured; no fixed accuracy guarantee is advertised.
Data handling arrangements
- Agree confidentiality terms before receiving confidential data.
- Use client-approved tools and transfer channels.
- Define authorised access, storage locations and retention/deletion terms.
- Obtain client approval before involving additional contributors.
- Do not upload client data to public AI tools without explicit permission.
Please describe your security and data-location requirements first. Do not send sensitive datasets, identity documents or confidential samples through the website assistant or an initial enquiry. This page does not claim ISO, SOC 2 or other security certification.
VENDOR ONBOARDING
A registered company.
A defined engagement.
Relevant verification documents can be provided privately to verified prospective clients through an agreed channel.
Company information
IA Data Solutions (SMC-Private) Limited
Incorporated in Pakistan
SECP Corporate Registration No. 0361082
Certificate of Incorporation and company FBR registration are available on request. Additional due diligence requirements can be reviewed before engagement.
Before work begins
- Agree scope, deliverables, acceptance and revision limits.
- Confirm pricing, payment milestones and delivery dates.
- Agree confidentiality, data rights and permitted contributor access.
- Confirm the company payment route before contracting.
Team capacity is assessed for each enquiry. No permanent team size or specialist certification is claimed.
COMMON QUESTIONS
Before your first pilot.
Can you quote before seeing the data?
A reliable quotation needs a sample or clear specification, volume, tool, quality requirements and deadline. Rates and turnaround are confirmed in writing after review.
Can we use our own annotation platform?
Share the platform and access requirements. We will review compatibility and onboarding needs before confirming the engagement.
Can you sign our NDA and complete vendor onboarding?
Send your proposed terms and vendor checklist for review. Acceptance is subject to reviewing the requirements; no agreement is implied by this page.
Can you scale beyond the pilot?
We can assess the required skills and availability after pilot review. Additional contributors require agreed access and confidentiality arrangements and client approval.
START A CONVERSATION
Tell us what your dataset needs.
Include the data type, approximate volume, guidelines, output format, quality expectations and deadline. We will review the requirements before proposing a paid pilot.
Request a Pilot Quotation ↗Opens a prepared email in your email application. Review it and press Send to contact us.