Image and Data Annotation for the Smart City

Smart City Annotation for Urban

Infrastructure and Mobility

Reliable training datasets for computer vision in smarter cities

What We Do

Precision Data Annotation for Urban Mobility and Smart Infrastructure.

We annotate images, video and multi-sensor data to train computer vision models for smart city applications. From traffic-flow analysis and road-user detection to road asset inventories and infrastructure condition assessment, our teams deliver reliable datasets for urban mobility and infrastructure management.

Rigorous and Controlled Process

Every project starts with a review of your data and annotation guidelines, followed by a pilot to validate labelling rules and quality criteria. Our teams then manage production, multi-stage quality checks and regular reporting on volumes, non-conformities and annotation traceability.

Security & Compliance (ISO 27001 / GDPR)

ISO 27001 : 2022-certified Infoscribe protects your data through controlled access, encrypted transfers and segregated environments. Depending on project requirements, we can anonymise faces and licence plates or work directly within your infrastructure to keep sensitive data under your control.

Multi-Sensor Annotation

We annotate images and video from fixed and vehicle-mounted cameras, LiDAR point clouds and geospatial data. Our capabilities include 2D/3D object detection and segmentation, tracking, temporal annotation and cross-sensor object association, all aligned with your urban data taxonomy and labelling rules.

EXpertise

Annotated Data for Urban Mobility, Public Safety and Infrastructure Management.

Infoscribe helps smart city technology providers prepare and annotate data for computer vision model training. Our teams combine human expertise, robust workflows and quality control to deliver consistent, production-ready datasets.

From street-level imagery and traffic video to 3D point clouds and geospatial data, we label the information needed for road-user detection, mobility-flow analysis, urban asset inventories and road infrastructure condition assessment.

2D & 3D Annotation for Computer Vision

Urban Safety and Incident Prevention

Image and video annotation for computer vision models used in public-space safety. We label and track people, abandoned objects, restricted-area crossings and unusual events according to client-defined criteria. Our teams deliver quality-controlled datasets for intelligent video analytics and incident-prevention applications.

Multimodal Mobility and Traffic Monitoring

Image and video annotation for road-user counting, classification and tracking, including cars, motorcycles, buses, trucks, vans, cyclists and pedestrians. Our teams work with toll-road camera footage and can label detailed vehicle attributes such as trailers, loads, roof-mounted equipment and specialised vehicle categories.

Traffic Management and Flow Optimisation

We annotate footage from intersections and road networks, including object detection, classification, segmentation and multi-object tracking for vehicles and other road users. These datasets help clients train AI models to measure traffic flows, assess lane occupancy and support smarter traffic management.

Smart Parking

We annotate images and video from car parks and on-street parking areas to detect and classify vehicles, identify occupied and available spaces, segment parking bays and track vehicle movements. Our datasets support computer vision models for parking occupancy analysis and management.

Waste Management and Intelligent Sorting

We annotate and segment images for computer vision models used in waste sorting and material-flow management. Our teams identify and outline objects in bins or on sorting lines, including overlapping, partially obscured objects and challenging lighting conditions, to create accurate datasets for waste recognition and classification.

Weather Monitoring and Sky Image Analysis

Semantic segmentation of wide-angle (fisheye) sky images to distinguish clear sky, clouds and aircraft contrails. Pixel-level annotation across changing weather and lighting conditions provides training data for atmospheric observation models.

Infrastructure Inspection and Maintenance

2D and 3D annotation to identify and outline defects in urban infrastructure, including cracks, deformation, corrosion, damaged signage and overgrown vegetation. We create datasets for visual inspection and predictive maintenance models.

Rooftop Chimney and Solar Panel Detection

We annotate aerial imagery and rooftop photographs to locate chimneys, solar panels and other building features. Object detection, bounding boxes and segmentation produce datasets for urban mapping, solar asset inventories and building stock analysis.

Crowd Counting and Public Event Analytics

Video annotation from fixed and mobile cameras to identify, count and track people at demonstrations, concerts, stadiums and pedestrian areas. The resulting datasets support crowd estimation and movement-flow analysis models.

Public Transport Inspection

Image and video annotation for buses, trains and trams to identify components and label visible defects such as cracks, impact damage and wear. We segment and classify anomalies to create datasets for automated visual inspection and maintenance.

Road Signs and Streetscape Analysis

2D and 3D annotation of traffic signs, signals, road markings, kerbs and roadside assets. We detect, segment and classify features using your taxonomy to train road-scene understanding and infrastructure mapping models.

Urban Energy and Street Lighting

Thermal and infrared image annotation to identify heat-loss areas, lighting equipment and visible defects. We outline and classify features according to your guidelines to support energy analysis and urban infrastructure inspection models.

Urban Waste and Street Cleanliness

Image annotation and segmentation to identify litter, abandoned objects, illegal dumping and waste collection assets. Our teams also label overlapping and partially occluded objects to build datasets for automated sorting and urban waste management.

Maritime and Port Monitoring

Image, video and LiDAR point-cloud annotation to identify and track vessels, boats, containers and port equipment. Detection, segmentation and multi-object tracking support computer vision training datasets for maritime traffic monitoring and port infrastructure safety.

Person Re-Identification Across Cameras

We review and match cropped images of people captured by different cameras, from different angles and at different locations. Using client-defined criteria, annotators assess whether images show the same person, creating reference datasets for training and evaluating person re-identification (Person Re-ID) models.

Urban Planning and Dynamic Mapping

2D and 3D annotation of satellite, aerial and mobile mapping imagery to identify buildings, roads, public spaces and urban assets. Segmentation, classification and feature delineation support automated mapping, urban modelling and change monitoring.

Text Data Annotation for AI

Incident Management and Maintenance

We annotate support tickets, technical reports and field service records to classify incidents, identify affected assets and extract information needed to manage and resolve issues.

Mobility Feedback and Citizen Engagement

We label complaints, citizen submissions and messages from participation platforms to categorise requests, identify topics and capture the needs expressed by residents.

Public Procurement and Tender Analysis

We annotate procurement and contract documents to extract selection criteria, technical requirements, deadlines, contract values and supplier award information.

Urban Planning and Sustainable Development

We annotate research reports, planning documents and public submissions to identify projects, locations, infrastructure and environmental priorities.

Autres Secteurs d'activité

FAQ

Frequently Asked Questions

Infoscribe AI works with companies and public-sector organisations developing smart city AI solutions, providing annotated data for model training and evaluation. We support a wide range of data types:

  • Images and video: traffic cameras, toll plazas, parking facilities, urban surveillance and vehicle-mounted cameras.
  • 3D LiDAR point clouds: object and infrastructure annotation, segmentation and classification.
  • Aerial and satellite imagery: buildings, roads, urban assets and solar installations.
  • Thermal and infrared imagery: equipment and visible anomaly annotation.
  • Geospatial data: labelling georeferenced features and linking annotations to mapping information.
  • Text documents: document classification, information extraction and structuring of technical reports, incident tickets and administrative records.

We also handle multi-sensor datasets while preserving supplied metadata such as timestamps, geographic coordinates and camera identifiers.

Each project follows client-defined labelling guidelines, classes and delivery formats, producing datasets ready for use in AI training and evaluation.

We support companies building computer vision solutions for urban mobility, public safety, infrastructure management and city services. Our teams prepare the annotated datasets needed to train and evaluate their AI models. Key use cases include:

  • Traffic and mobility: vehicle counting and classification, pedestrian and cyclist tracking, traffic flow and intersection analysis.
  • Roads and infrastructure: asset inventories, defect identification, traffic sign and road marking annotation.
  • Smart parking: vehicle identification, parking bay annotation and occupancy classification.
  • Urban safety: video event annotation, person and object tracking, and cross-camera visual re-identification (Person Re-ID).
  • Waste management: object segmentation and classification for sorting and street cleanliness applications.
  • Mapping and urban planning: identifying buildings, chimneys, solar panels and urban assets in aerial or satellite imagery.
  • Environment and weather: cloud-cover segmentation and atmospheric observation imagery.
  • Transport and urban logistics: public transport vehicles, port equipment and related infrastructure.

We support both pilots and large-scale production, with project-specific guidelines and quality criteria.

Infoscribe AI specialises in data annotation; our clients remain responsible for developing and operating their computer vision models.

Infoscribe AI annotates images and video used to train computer vision models for road-user counting, classification and tracking. We work with footage from toll plazas, intersections, road networks and traffic-monitoring cameras.

  • Vehicle detection and classification: cars, motorcycles, buses, heavy goods vehicles, vans, camper vans and client-specific categories.
  • Vulnerable road users: pedestrians, cyclists and powered two-wheeler riders.
  • Detailed vehicle attributes: trailers, loads, roof equipment, special markings and other visible features.
  • Multi-object tracking: linking the same vehicle or road user across successive frames to capture trajectories.
  • Traffic events and interactions: restricted-area crossings, lane changes and other events defined in the annotation guidelines.

Depending on model requirements, we deliver bounding boxes, polygons, segmentation masks or trajectories.

Our annotation guidelines define classes, labelling conventions and how to handle occlusions, partially visible vehicles and complex traffic scenes. The resulting datasets help clients train and evaluate traffic-counting, vehicle-classification and mobility-analysis models; we do not develop those algorithms.

Every Infoscribe AI project is designed around your data, model architecture and training or evaluation objectives. We adapt our annotation workflow to your technical specifications:

  • Annotation methods: 2D/3D bounding boxes, polygons, semantic and instance segmentation, classification, multi-object tracking and temporal labelling.
  • Classes and ontologies: your taxonomy, attributes, class hierarchies and rules for ambiguous or partially visible objects.
  • Delivery formats: COCO, YOLO, JSON and custom structures compatible with your training pipelines.
  • Tools and environments: our annotation platforms or your own tools, including CVAT, Label Studio and specialist 3D software.
  • Quality requirements: acceptance criteria, tolerances and checks matched to the annotation task.

We generally recommend a pilot on representative data to validate the guidelines, surface edge cases and confirm annotation accuracy before scaling up.

Our teams can work with your computer vision engineers as requirements evolve, while maintaining dataset consistency and traceability.

Yes. We support computer vision projects combining footage from multiple cameras, LiDAR sensors and other acquisition systems. Our annotation services include:

  • Multi-camera tracking: matching the same object or vehicle across video sequences using available information and client-defined rules.
  • Person re-identification (Person Re-ID): comparing cropped images taken from different angles, times or locations to assess whether they show the same person.
  • Multi-sensor annotation: labelling objects in 2D images and 3D point clouds and matching annotations where data is synchronised and calibrated.
  • Temporal annotation: tracking objects and events over time while retaining supplied identifiers and timestamps.

We account for viewpoint changes, occlusions, lighting variation and image quality.

For Person Re-ID, we apply documented comparison criteria and can mark cases as inconclusive when the visual evidence is insufficient. These labels support model training and evaluation; they do not establish a person’s identity with certainty.

Quality at Infoscribe AI is built on a structured workflow, project-specific training and multiple levels of review. Our process covers:

  • Requirements review: assess data, guidelines, classes and acceptance criteria.
  • Pilot annotation: validate rules, identify edge cases and align annotator decisions.
  • Training and calibration: prepare production teams using the approved guidelines and pilot feedback.
  • Multi-stage quality control: dedicated reviewers, independent checks and project manager audits.
  • Monitoring and corrections: track non-conformities, review quality metrics and correct errors as required.

We offer contractual quality targets of 95%, 97% or 99%, depending on project requirements and budget. If the agreed threshold is not met, non-compliant annotations are corrected at no additional charge.

Measurement criteria are agreed with each client and adapted to the task, whether classification, detection, segmentation, tracking or 3D annotation. This helps maintain consistency across annotators, batches and production phases.

The EU AI Act sets requirements for certain high-risk AI systems, including provisions on the governance, quality and documentation of training, validation and testing datasets.

  • Define annotation rules: document classes, labelling criteria and how ambiguous cases are handled.
  • Document the process: retain guidelines, versions and decisions on edge cases.
  • Check data quality: assess accuracy, completeness and consistency against agreed criteria.
  • Flag anomalies and inconsistencies: identify ambiguous samples, labelling errors and issues affecting dataset quality.
  • Track corrections: record non-conformities and corrective actions under the agreed project workflow.

Infoscribe AI can therefore help clients implement documented, quality-controlled annotation processes relevant to the data governance requirements of Article 10 of the AI Act.

This may be relevant to smart city systems involving traffic management, certain critical infrastructure or surveillance, depending on the system’s intended purpose and regulatory classification.

Infoscribe AI does not certify AI Act compliance. Our services support data quality and documentation; overall compliance remains the responsibility of the relevant AI system stakeholders under their applicable legal obligations.

Infoscribe AI is ISO 27001:2022 certified and follows information security procedures suited to sensitive data, including surveillance footage, traffic video and material that may contain personal information.

  • Access control: restrict data access to authorised personnel with a project-related need.
  • Secure transfers: use encrypted protocols and file-integrity verification procedures.
  • Segregated environments: separate projects and manage access rights to reduce disclosure risks.
  • Traceability: log access and activities according to the tools and infrastructure used.
  • Contractual confidentiality: define confidentiality commitments and permitted use of client data.

We can also work directly within a client’s infrastructure. Our annotators securely connect to the client’s servers or annotation platform so the data remains in the client environment.

Where required, we can support anonymisation tasks such as masking faces and vehicle licence plates.

For personal data, processing arrangements are agreed contractually to address applicable GDPR requirements, including confidentiality, security and processor obligations. ISO 27001:2022 certification covers our information security management system; it is not, by itself, a certification of GDPR compliance.

Yes. Infoscribe AI can work directly in your annotation platform or provide its own tools and infrastructure, depending on technical and security requirements. Our teams use a broad range of software:

  • Image and video annotation: CVAT, Label Studio, RedBrick AI, Segments.ai and Picsellia.
  • 3D point clouds and LiDAR: CloudCompare and Sustech 3D.
  • Geospatial data and mapping: QGIS.
  • Proprietary platforms: our annotators can be trained on your internal tools and workflows.

Where confidentiality requires it, annotators can connect securely to your servers without moving datasets to our infrastructure.

With more than 300 annotators available for deployment, we can scale teams to match project volumes and deadlines. We support everything from pilots involving a few hundred images to large programmes requiring thousands of annotation hours per month.

Dedicated project managers and quality reviewers oversee scaling, consistency and contractual delivery commitments. This gives clients additional annotation capacity without the overhead of recruiting, training and supervising new in-house teams.

Each Infoscribe AI project follows a structured process designed to clarify requirements, deliver consistent annotations and meet agreed timelines:

  • Requirements discovery: review data, objectives, annotation classes, delivery formats and technical or confidentiality constraints, using representative samples and guidelines.
  • Scoping and quotation: assess complexity, volumes and staffing needs, then provide a commercial proposal covering pricing, timing and quality commitments.
  • Pilot phase: annotate a representative batch to validate instructions, measure processing times and identify ambiguous cases; incorporate client feedback before scaling.
  • Production and quality assurance: a dedicated team annotates the data under project management supervision, with staged checks and regular progress reviews.
  • Delivery and refinement: provide annotated data in the agreed formats with the contracted reporting; correct non-conformities in line with the quality commitment.

We support both one-off projects and ongoing engagements, with phased deliveries and flexible team sizes.

Our aim is to fit into your AI development cycle and supply reliable datasets ready for model training and evaluation.

The cost of a smart city data annotation project depends on the data type, annotation complexity, volume and required quality level. Key pricing factors include:

  • Data type: images, video, LiDAR point clouds, aerial imagery or text documents.
  • Annotation method: classification, bounding boxes, segmentation, multi-object tracking or 3D annotation.
  • Scene density and complexity: object counts, occlusions, contour precision and class diversity.
  • Project volume: number of images, video duration, objects or 3D data volume.
  • Quality and delivery requirements: review levels, deadlines and resources.

Infoscribe AI prepares a tailored quote using your specifications and representative sample data. Pricing may be per image, per object, per frame or per production hour, depending on the task.

When needed, a pilot helps measure annotation time, validate instructions and refine the budget before production begins. Volume-based arrangements are available for recurring or large-scale work.

Send us a sample dataset and your annotation guidelines to request a quote for your smart city project.