October 2026 Summaries

Encord’s October 2026 product updates focus on accelerating the conversion of raw robotics, sensor, and medical-imaging data into training-ready datasets. New beta support lets teams upload and label MCAP recordings directly on their original timestamps, visualize asynchronous sensor streams on a unified timeline, and annotate time-series signals alongside camera and LiDAR data. Workflow-stage class controls allow organizations to restrict which annotation classes individual annotators and reviewers can edit or view, helping specialists focus on relevant tasks and protecting completed work. Medical-imaging improvements enhance DICOM and NIfTI visualization, including oblique and multiparametric views, 4D NIfTI labeling, density projections, faster rendering, metadata and embedding-based search, task preloading, and cloud synchronization for NIfTI files. Additional updates include polygon splitting and vertex snapping, simplified 3D cuboid interpolation, a redesigned issue-review interface, and signed task-ready notifications for Custom Agents using the updated SDK.

Encord has introduced beta native support for MCAP, an open robotics recording format that stores synchronized yet independently timestamped data from sources such as cameras, LiDAR, IMUs, joint states, and transforms. The feature is intended for robotics and physical AI teams that need to annotate multi-sensor recordings without converting files or forcing data onto a common frame grid, which can inflate frame counts, miss short events, and create timestamp offsets. Users can upload MCAP files directly through the platform, SDK, or cloud storage; view visual, point-cloud, and numeric streams together on their original timeline; annotate 3D objects and time-series intervals; control playback snapping; and export labels tied to nanosecond-level recording timestamps. Available to Encord platform customers, the beta initially supports 3D and time-series annotation and is being developed with feedback from beta users.