FiftyOne Search
FiftyOne Search makes multimodal datasets searchable. Query by natural language, visual similarity, or metadata to inspect model behavior in context, isolate failures, and find related scenarios for training.
FiftyOne Features
Search, visualize, and debug multimodal datasets
FiftyOne brings search and visualization into one workflow. Stop writing custom scripts and jumping between tools to answer questions about your data.
Search your entire dataset
FiftyOne creates a persistent dataset index so natural language, visual similarity, and metadata queries stay fast and interactive. Surface related samples, episodes, and behaviors across your entire multimodal dataset.
Visualize in context
Explore episodes through synchronized playback that aligns every modality on a shared timeline. Customize tiles to inspect images, 3D scenes, sensor streams, plots, maps, and annotations together.
Search & Query
Find the episodes that matter
Search to find relevant episodes across your entire dataset. Powered by state-of-the-art embeddings and a persistent dataset index for fast, interactive search.
Generate embeddings for individual segments—not just whole episodes—to retrieve similar results within any filtered subset of your dataset.
Describe what you're looking for in natural language with FiftyOne Agent. Search for objects, scenes, behaviors, or failure modes without writing filters or queries.
Define custom events from multimodal sensor data, then search for every matching episode.
Find semantically similar data across your dataset without requiring labels or metadata.
Filter your dataset by labels, tags, scalar fields, and custom metadata to isolate specific scenarios or cohorts.
Explore embeddings-based workflows for robotics and AV
Learn how embeddings reveal the structure of complex multimodal datasets across robotics, AV, and other physical AI workloads. Explore practical workflows for search, curation, evaluation, and model improvement.
Visualization
Visualize your data in context
Explore multimodal data in context to uncover patterns, quality issues, and gaps that are hard to find in aggregate metrics.
See every modality in sync
Sync multi-rate sensor data to a shared timeline. Images, 3D scenes, plots, sensor streams, and metadata stay aligned as you scrub episodes.
Browse datasets at scale
Quickly scan thousands of samples or episodes on the grid.
Configure data overlays
Toggle labels, predictions, metadata, and custom fields to see exactly the information you need.
Group related views
Organize and link multi-view samples with grouped datasets. Groups don't require a shared timeline.
Trace metrics back to data
Every visualization in FiftyOne is interactive, letting you inspect underlying samples with a single click.
Physical AI Search
Search robotics and AV episodes
Improve VLA generalization with less data
Instead of burning budget on massive, redundant datasets, FiftyOne isolates the high-value, diverse demonstrations your vision-language-action (VLA) model needs to generalize better.
Search petabyte-scale data for the long tail
As data scales and standard scenarios yield diminishing returns, FiftyOne makes massive datasets searchable, empowering autonomous vehicle teams to target longtail investigations and surface rare edge cases.
Get started with FiftyOne Enterprise
Questions?
We have answers.
Yes, FiftyOne supports multimodal data. FiftyOne is a unified data platform for physical AI that brings visualization, curation, and annotation of multimodal, time-series data into one place. Teams inspect synchronized camera, LiDAR, sensor, and robot state data, then search and curate it at scale to build better datasets for autonomous vehicles and robotics.
FiftyOne natively ingests MCAP, the default recording format for ROS 2, and works with data converted from rosbag. You can also bring your own extracted frames and post-processed sensor data.
Camera streams, LiDAR point clouds, and numeric time-series sensors like IMU, acceleration, and CAN bus signals, along with robot state, all play back in sync on a single shared timeline. Every modality stays aligned, so you can inspect a complete episode in context rather than one sensor at a time.
No. FiftyOne Multimodal works directly with encoded stream media and infers frame-level data from it, so you do not have to pre-extract frames before ingesting. If your pipeline already produces extracted, post-processed frames, you can bring those too.
Visualization tools make it easy to inspect a single robot episode or failure. FiftyOne Multimodal adds what those tools do not: large-scale curation, embeddings and semantic search, and annotation, so you can go from finding an edge case to a labeled, model-ready dataset in one platform.
You can query by metadata, temporal events, and annotation labels, then search by visual and semantic similarity using embeddings. This surfaces rare scenarios, recurring failure modes, and underrepresented data across an entire dataset, the same longtail mining workflow autonomous vehicle teams rely on.
Yes. You can tag arbitrary time intervals to curate segments for investigation, work with 2D bounding box and 3D cuboid labels at the message level, and view labels as tracks over time. Label tracks show exactly when an object appears or an event occurs across a full scene.
It is built for datasets ranging from thousands of scenes to hundreds of millions of frames, with support for long, multi-hour encoded video and an open-ended number of time-series sensors. Purpose-built indexing keeps search, filtering, and playback fast at that scale.
© 2026 Voxel51 All Rights Reserved