Sycrad's roadmap expands the product without compromising native performance, local data ownership, or the ability to review every result. These are active product directions rather than fixed release commitments. Their integration order and delivery timelines have not yet been set, and priorities may change as pre-alpha testing reveals where the greatest practical gains are.
Collaboration system
Sycrad currently supports one annotator per project. The Team plan will move collaboration into the application, coordinating cuboid metadata through Sycrad's service while point cloud data remains on each user's machine. Teams will be able to see who is working in a scene, follow audit history, and route cuboids that need review. Enterprise deployments will provide the same workflow through a customer-controlled server on a private network.
Panoptic segmentation
Panoptic segmentation unifies categorical semantic labeling with instance-level identification. While Sycrad currently focuses on 3D bounding cuboids, future workflows will introduce panoptic labeling to partition point sweeps into amorphous background classes (such as road surfaces, terrain, and building facades) and discrete foreground objects (such as individual vehicles, pedestrians, and cyclists) with persistent instance IDs. This delivers full structural scene context alongside isolated object instances in a unified workflow.
Gaussian Splatting scene reconstruction
Integrating 3D Gaussian Splatting (3DGS) will fuse multi-view calibrated camera imagery with registered LiDAR sweeps into photorealistic 3D scene reconstructions. Operators will be able to inspect complex scenes against real-world photometric geometry directly inside the viewport. Leveraging Sycrad's SLAM pipeline experience in isolating dynamic returns, moving actors will be handled to prevent motion blur and ghosting artefacts during reconstruction, delivering clean visual context across dynamic sequences.
AI-ready export pipeline
The planned export pipeline will deliver structured labels for AI perception model training, eliminating the need for secondary data parsing or normalization scripts before ingestion. Annotated sequences will export with cuboids, segmentations, tracking identities, motion states, situational reasoning attributes, and camera projections.
Adverse weather
Rain, snow, and fog introduce returns that obscure real geometry and slow down annotation. Planned work will investigate ways to identify and suppress these weather artefacts without discarding useful scene structure, giving annotators a cleaner starting point in difficult captures.
Continued SLAM development
SLAM will remain an active area of development. The work includes broader dataset coverage, stronger dynamic point detection, robust DBSCAN clustering, multi-frame tracking, and refined ground segmentation across diverse sensor layouts and complex environments. Each stage will continue to be measured independently against ground truth so improvements remain attributable and reviewable. Learn more about current pipeline mechanics in our FAQ.
AI-assisted labelling
Future auto-labelling workflows are planned around two deployment paths: local inference for data-sensitive environments and server-backed inference where shared infrastructure is appropriate. Both paths are intended to accelerate annotation while keeping generated labels explicit and reviewable inside Sycrad.
Scripting support
Team and Enterprise plans are planned to gain a controlled scripting interface for repeatable project tasks, workflow automation, and integration with internal data pipelines. The goal is to make large annotation operations reproducible without turning routine automation into manual UI work.
Broader dataset support
Supported dataset coverage will progressively expand across development phases. Annotation workflows are currently under development for PCD and LAS/LAZ formats. Autonomous vehicle coverage will broaden to include formats such as Argoverse. While Sycrad does not currently support robotics datasets, future updates will introduce specialized robotics data support, extending interpolation and SLAM workflows beyond autonomous vehicles. Format support will subsequently expand to accommodate aerial survey, defense reconnaissance, and geospatial mapping pipelines. Sycrad is also planned to read datasets directly from cloud storage, reducing the need to stage every source locally before work begins.
Linux support
macOS and Windows remain the primary platforms through pre-alpha and the path to the first general release. Native Linux (Ubuntu/Debian) desktop builds are planned once the core desktop workflow, rendering pipelines, and release infrastructure have stabilized. For current operating system and rendering backend requirements, see our FAQ.