CASE STUDIES

Case studies

From a data brief.To a traceable delivery.

Explore scene planning, challenges, methods and handover through three data projects.

Browse case studies Explore delivery & quality

PROJECT LIBRARY

Explore projects by task.

Choose a task, read its brief and open the full case for scope, methods and review records.

Showing 3 projects

Image and point-cloud data association — DATA SERVICES · PROJECT SCENARIOS

AutomotiveProject in focus

One road scene. Connected sensor data.

Turn complex road scenarios into inspectable batches, linking images, point clouds and object identities.

6 cameras + 1 LiDAR
2D / 3D associations
Scene coverage

sequences: 12
Synchronized keyframe groups: 1,200
Scene types: 4

Read the full case Explore automotive solutions

Read the project briefClose the project briefAUTO-001

A perception team wants to examine multimodal associations and challenging scenes in one organized batch.

Project need: Organize 12 sequences and 1,200 synchronized keyframe groups. Each group links six images and one point cloud across junctions, night and backlight, roadworks and mixed traffic.
Key challenge: Occlusion and low light make boundaries uncertain. The same object across cameras and LiDAR needs consistent identity, timing and coordinate references.
Approach: Calibrate rules using edge cases, then review objects, tracks and 2D / 3D links by sequence. Associate issues with scenes and rule versions for revision.
Handover: Sequence indexes, frame labels, cross-modal links, calibration references and batch-review records.

Discuss a perception task

Road-data collection vehicle and sensors — DATA SERVICES · PROJECT SCENARIOS

Data collectionCOL-001

From a scene gap to a pilot plan.

Define a road-video batch through scene groups and usability requirements.

Road video
Scene planning
Pilot review

Clips · target: 120
Per clip: 20 sec
Scene groups: 3

Read the full case Explore data collection

Read the project briefClose the project briefCOL-001

A data team needs daytime, nighttime and suburban road scenes, with a plan that goes beyond volume alone.

Project need: Plan 120 clips of 20 seconds each: 60 urban daytime, 30 urban nighttime and 30 suburban daytime. Collection locations and equipment are agreed separately in the project plan.
Key challenge: A full clip count does not ensure full coverage. Duplicates, low-light usability and distant objects need separate checks and scene records.
Approach: Define scenes and capture conditions, validate file requirements with a pilot, then review counts, decoding, duplicates and coverage gaps by group.
Handover: Clips, file indexes, scene tags, collection guidelines and issue lists. The collection plan remains pending; pilot review and production execution are tracked separately.

Discuss a collection task

Bounding-box annotations in a road image — DATA SERVICES · PROJECT SCENARIOS

Data annotationANN-002

A small image batch. A shared rulebook.

Start with class definitions and object boundaries to build consistent annotation instructions.

2D bounding boxes
Class definitions
Batch review

Images · scope: 360
Object classes: 4
Pilot batch: 1

Read the full case Explore data annotation

Read the project briefClose the project briefANN-002

A vision team wants to clarify classes, occlusion and truncation before scaling production.

Project need: A 360-image pilot covers vehicle, pedestrian, cyclist and traffic-sign classes. Scope and review status are tracked by image, with bounding-box counts recorded separately.
Key challenge: Small, partially visible and truncated objects can produce different interpretations. Class names alone are not sufficient production guidance.
Approach: Collect boundary examples and exclusions, annotate and review a pilot, then freeze the rule version. Record disagreements, revisions and linked samples.
Handover: Image indexes, bounding-box labels, class definitions, rule versions and issue records. Confirm formats and acceptance criteria using actual samples.

Discuss an annotation task

FROM BRIEF TO DELIVERY

A related task. Your own requirements.

Use these cases as a starting point. Define volume, task boundaries and acceptance criteria around your samples and objectives.

01
Task & scenarios
What should the model do? Which scenes, classes and data modalities matter?
02
Existing data & gaps
Prepare de-identified samples, file structures and known issues to identify collection or annotation needs.
03
Expected handover
Describe usage, preferred formats and review requirements, then use a pilot to confirm the instructions.

PROJECT BRIEFS

Bring a clear structure to the discussion.

Download scope, scene and deliverable notes for all three projects to understand their structure and working methods.

Download project briefsJSON · Project documentation · No sign-in

case-study.json

{
  "sourceStatus": "project-record",
  "projectCount": 3,
  "projectIds": [
    "AUTO-001",
    "COL-001",
    "ANN-002"
  ]
}

Structure excerpt · full file contains project objects and scope fields

How is a delivery checked?Explore pilots, production review, issue resolution and batch acceptance.Explore delivery & quality

QUESTIONS & ANSWERS

About our case studies

01Can we commission just one part of the work?

Yes. Teams with existing data, guidelines or platforms can define a collection, annotation, review or delivery-preparation scope. Each case includes a relevant service and inquiry link.

02What do the downloads include?

The files contain project scope, scene allocations, deliverable types and review summaries. Source images, video, point clouds and labels are delivered separately under project agreements.

03Can a case scope be used as a quote?

Cost and timing depend on samples, scene coverage, task complexity, quality criteria and delivery requirements. A pilot helps establish the execution scope.

04How do we discuss a similar project?

Share your target scenes, data types, existing samples and expected volume through project inquiry to define a suitable pilot and delivery plan.

LET’S DEFINE YOUR PROJECT

Start the next brief with your task.

Share target scenes, existing data and current challenges to shape a reviewable scope and pilot plan.

Discuss your project