Vieva AI | AI Data from Vietnam
Start a Project

Good AI Starts With a Clear Standard.

The robot hand must be in frame.

A bare robotic hand with its wrist and forearm
Fingertips, hand, gripper, or the whole wrist?
A robotic hand covered by a white glove
Does a gloved hand count?
A robotic hand cut off by the left edge of a wide-angle photo
Which camera? Full image or a crop?
A robotic hand holding a large metal tool that hides part of the hand
Partly in view, or fully? Is it still valid if hidden?
Illustrative criteria

Object Definition

A robotic hand includes the end-effector or gripping mechanism, with explicitly defined boundaries for fingertips, grippers, hand assemblies, and wrist components.

Frame Boundaries

Specify whether the entire target object must remain within the image boundaries, whether partial inclusion is acceptable, and whether any minimum spatial margin is required.

Visibility and Occlusion

Establish the minimum required visibility of the target object and determine whether partial occlusion by other objects is acceptable.

Special Cases and Exceptions

Specify how exceptional conditions, such as gloves, unusual gripper designs, visual obstruction, camera distortion, or uncertain object boundaries, should be handled.

What does your requirement mean in practice?
Have you defined your requirements?
  • Shared criteriaWhat counts, agreed before work starts.
  • Practical guidelinesRules two people read the same way.
  • Examples your team can useCorrect, incorrect and edge cases.
Explore Our Evaluation Services

Evaluate the delta, elevate the data

Small gaps in requirements can become big differences in AI data

Data & Tasks We Support

Tell us what you’re working with. Many projects combine more than one of these.

  • Speech & Conversations
  • Documents & Handwriting
  • Image & Video
  • Computer-use & Agent Trajectories
  • Vietnamese Language Tasks
  • Preference Data & Human Evaluation

Start with a pilot.

Validate the approach, review the data, and align on quality before scaling.

  1. Discuss the project
  2. Agree on the approach
  3. Deliver and review
  4. Scale when it fits

The team behind the data

Vieva is based in Hanoi and serves international AI teams.

3,000+computer-use trajectories with synchronized action logs
2,500+corrected agent trajectories, with task completion up about 30–35%
1,500 hoursa month of ego-perspective video for robot learning

Tell us what you’re building.

Share the project, the data type, and your expected timeline.

Start a Project