Label objects and regions in 3D.
Annotate 3D bounding boxes, semantic classes, and instances in point clouds, with linked camera views to help review the scene.
- 3D object detection
- Semantic and instance segmentation
- Linked camera views
Workflows
Choose the right workflow for every task
Annotate, review, and export point cloud, image, document, video, and multimodal data in one workspace, with model assistance when you need it.
Label the image, the object, or both.
Build multiclass, multilabel, and hierarchical classification datasets without separating image-level decisions from shape-level review.
- Image and shape labels
- Reusable class groups
- Dataset statistics
Turn dense pages into editable structure.
Parse text, tables, formulas, and document layouts with PaddleOCR, then inspect and correct every region in the same workspace.
- Text recognition
- Document layout
- Key information extraction
Describe what changes across the timeline.
Create and review labeled time segments, navigate long clips efficiently, and preserve the temporal context behind every classification.
- Timeline segments
- Frame-accurate review
- AI-assisted descriptions
Build structured answers from visual evidence.
Design repeatable question templates, combine manual and model-assisted answers, and export data for modern vision-language training.
- Question templates
- Model-assisted answers
- Structured VQA datasets
Turn an image into a reviewable conversation.
Talk to a vision-language model with the current image in context, refine the response, and preserve approved conversations as training data.
- Context-aware conversations
- Image captioning
- ShareGPT export