Bounding boxes
Object extents for detection tasks, with agreed rules for truncation, overlap and partially visible objects.
COMPUTER VISION / ANNOTATION
Image & video
annotation.
Image and video annotation guided by explicit object definitions, boundary rules and frame-to-frame consistency. Start with a sample that makes the requirements concrete.
12+ years of relevant professional experience · Direct founder involvement
IMAGE & VIDEO CAPABILITIES
Labels aligned with
your vision task.
Annotation translates a visual specification into structured observations. Clear taxonomies and explicit edge-case rules help make labels easier to review and use.
Polygons & semantic segmentation
Object outlines and class regions, with boundary and inclusion rules defined for the task.
Keypoints & landmarks
Points following a specified schema, with visibility and ambiguity handling agreed in advance.
Object tracking
Persistent identity across selected frames, with review of re-entry, occlusion and identity changes.
Frame consistency
Checks for label drift, missed objects and inconsistent interpretation across a sequence.
Guideline calibration
A representative sample to clarify classes and edge cases before committing to a larger scope.
SYNTHETIC WORKSPACE
Different labels.
One consistent specification.
Switch between illustrative annotation types. The vector scene is original; it contains no client imagery and no measured performance results.
ORIGINAL SYNTHETIC SCENE · ILLUSTRATIVE ANNOTATIONS
Rules make labels useful.
Box placement follows the agreed object-inclusion rules. This example identifies vehicles and pedestrians.
FRAME AVEHICLE_01
FRAME BVEHICLE_01
Simplified frame pair: the same ID follows the same object. Frame-to-frame review also checks visibility, class and boundary changes.
WHAT YOU RECEIVE
Outputs with context.
Final scope, tools and export requirements are confirmed before work starts. The items below describe proposed deliverable types, not a promise of every format or integration.
Annotated sample or agreed batch
Labels prepared to the agreed taxonomy. Required tools and export compatibility are assessed before acceptance.
Guideline clarifications
Decisions on ambiguous objects, visibility, boundaries and class definitions recorded for review.
Quality review findings
Observations, flagged cases and correction notes for the agreed review scope.
QUALITY BY DESIGN
Agree on quality.
Then evaluate the work.
A focused pilot helps establish the task rules, review method and acceptance criteria. No generic accuracy promise replaces a task-specific agreement.
Specify
Agree on classes, object boundaries, visibility flags and sequence rules.
Pilot
Annotate a representative sample with the proposed specification.
Review
Resolve ambiguous cases and inspect labels against agreed criteria.
Deliver
Provide the agreed annotations and supporting review notes.
PROJECT QUESTIONS
Clarity before
commitment.
Scope and suitability are reviewed directly with the founder.
Can you use our existing annotation guidelines?
Yes, the initial discussion can start from your guidelines. A pilot helps identify ambiguities and assess whether the task, tools and timeline are a suitable fit.
Which tools and export formats are supported?
Relevant experience includes CVAT and Roboflow. Specific access, workflow and export requirements must be checked before a delivery commitment; universal format support is not promised.
How are difficult or partially visible objects handled?
The specification should state whether to include visible extent, inferred extent or a flag for uncertainty. These choices are agreed for the task rather than assumed.
Can you handle large video volumes?
Volume, frame sampling, annotation complexity and review needs are assessed against current capacity. Altema does not claim an in-house production workforce or unlimited throughput.
START WITH A CLEAR SCOPE
Bring the challenge.
Let’s define the pilot.
Independent, founder-led delivery. Capacity, timing and requirements are assessed for each project.