Segments.ai

Segments.ai operates a multi-sensor data labeling platform used by computer vision and machine learning teams at robotics and automotive companies to annotate 3D point cloud (lidar) and 2D image data for AI training and validation. The platform allows several sensors to be labeled at the same time, maintains consistent track IDs across modalities and frames, and can project or copy labels from 3D sensors onto corresponding 2D images so that a single cuboid replaces many manually drawn bounding boxes. A fused view shows point clouds overlaid on camera images to give labelers additional context in noisy scenes.

For 3D work, the product supports point clouds of unlimited size by splitting them into 3D tiles to reduce loading time and GPU load, a Batch Mode that shows every instance of an object in a sequence on one scrollable screen so a single edit propagates across frames, a Merged Point Cloud mode that combines frames so sparse stationary objects can be annotated with a tight cuboid in one pass, keyframe interpolation, automated cuboid tracking with manual oversight, and full three-axis cuboid rotation. For 2D images and sequences, machine-learning-assisted tools provide segmentation, bounding boxes, polygons, polylines and keypoints, including a Superpixel 2.0 tool that divides an image into coherent regions using shape-aware models tuned for autonomous driving and robotics, plus autosegment, hover-and-click and pre-label workflows.

The company also maintains a public Python SDK, documentation, a Voxel51 plugin and open-source research repositories (including work combining Segment Anything with Grounded DINO and CLIPSeg, and a latent diffusion approach to panoptic segmentation presented at ECCV 2024).

Founding story

The company was founded in early 2020 by Otto Debals (CEO) and Bert De Brabandere (CTO) after both completed PhDs and gained industry experience, including in automotive. Debals holds a PhD in signal separation/multilinear algebra from KU Leuven and previously spent two years at McKinsey; before Segments.ai he and De Brabandere had together launched Investimize, a financial robo-advisor. De Brabandere holds a PhD in computer vision from KU Leuven, interned at Google DeepMind and won the Qualcomm Innovation Fellowship.

Business model

Software platform sold to engineering teams, with tiered plans (including Scale and Enterprise tiers that unlock automated tracking) and API/SDK integration into customers' data and model-training pipelines; outsourced labeling services are also available alongside the self-serve tooling.

Subscription-based access to the labeling platform across multiple pricing tiers ranging from startups to large enterprises, with optional outsourced labeling services.

Traction

Team size of 8 as of the Y Combinator profile, with the platform used by large organizations and startups building autonomous drones, delivery robots and self-driving cars. Public repositories include a Python SDK and research code with several hundred GitHub stars. Preqin reports EBITDA of -$32,344 and net income of -$39,429 for fiscal 2021 and EBITDA of -$29,245 and net income of -$30,498 for fiscal 2022.

Latest developments

Uber Technologies acquired Segments.ai BV in October 2025 for an undisclosed amount, with the stated aim of adding the robotics annotation platform and team to Uber's lidar and multi-sensor labeling tooling. The company's website and blog remained active into 2026, publishing monthly autonomy industry roundups.