Beyond visualization: query, annotate, and curate robotics data

Choosing the right infrastructure for your physical AI data is critical. Foxglove is a robotics data platform for collecting, visualizing, and debugging fleet data, with search across recordings and beta semantic search for curating training data. Voxel51 is a multimodal data platform that turns those same recordings into curated datasets, model evaluations, and annotated training data. Both read MCAP files and LiDAR point clouds, but they are built for different stages of the development lifecycle.

Book a demo Learn more about Voxel51

Voxel51 and Foxglove compared for physical AI and robotics data workflows

Foxglove Alternative

Why teams choose Voxel51 over Foxglove

Foxglove is built to get data off robots and help teams find and debug what happened. Voxel51 goes beyond visualization, giving teams the tools to understand model behavior, curate datasets, and annotate episodes across the physical AI lifecycle.

Move beyond inspection to model iteration

While Foxglove shows you what your robot did, Voxel51 transforms recorded episodes into gold-standard datasets to train better models.

Query at dataset-scale

Query sensor values, events, and labels across every episode with columnar Parquet projection tables (beta), then move results straight into curation, labeling, and evaluation.

Unified data platform for VLA development

Bring visualization, data understanding, curation, and multimodal annotation together into a single continuous physical AI data flywheel.

Voxel51 Features

Voxel51 turns raw sensor logs into your core model advantage

Stop drowning in unindexed MCAPs and wrestling with custom scripts just to isolate edge cases. Voxel51 helps teams understand model behavior at scale, turning multimodal datasets into better models, faster.

Query massive datasets by metadata, temporal events, and annotation labels.

Features

Foxglove vs. Voxel51 at a glance

Foxglove focuses on collecting, visualizing, and debugging fleet data, live or recorded. Voxel51 expands visualization into an end-to-end data platform for VLA and perception models.

Data modalities
Primary focus	Collecting, searching, visualizing, and debugging fleet data, with curation of training datasets	Turning data into better models: dataset curation, model evaluation, recorded log analysis
Core visualization (multi-sensor sync, timeline, playback)	Free plan, closed source	Free and open source in FiftyOne
Live vs. recorded	Both live and recorded data, with remote access and teleoperation	Reads MCAP as recorded data to curate datasets & evaluate models
VLA development support	Curates training datasets with sessions, events, and the agent. No VLA-specific tooling listed	✅ Supported. LeRobot v3 import and export, plus episode-level curation and temporal tags
Fleet-wide querying	Structured search across recordings, events, and sessions by device, time, event type, message content, and metadata, backed by a columnar index of decoded message fields	Columnar projection tables over indexed MCAP data (beta) scan only the columns a query needs
Embeddings and similarity search	Semantic search (beta, access by request) over image and video topics, using NVIDIA Cosmos video-text embeddings	✅ Supported. Segment embeddings for MCAP recordings are in beta
Scenario mining	Natural-language semantic search (beta), plus structured search on message content, events, and metadata. No dedicated scenario mining listed	✅ Supported via queryable Segment Embeddings and Events projections (beta)
Temporal tags & label tracks	Foxglove Events & Event Types. Acts as global triage tags across your entire data lake to bookmark system faults or disengagements.	Temporal Tags & Label Tracks. Acts as ML dataset assets that can be exported straight into a model training pipeline.
Annotation & auto-labeling	Event annotation on time ranges. No ML labeling or auto-labeling listed	✅ Supported
Model evaluation	Comparison Mode for side-by-side runs. No model evaluation metrics listed	✅ Supported
Versioning & dataset lineage	Not listed on foxglove.dev	✅ Supported
Data quality and label issue detection	Not listed on foxglove.dev	✅ Supported
Deployment	Foxglove Cloud, bring your own storage (AWS, GCS, Azure), an on-prem data plane in your own Kubernetes, edge sites, and fully offline deployments by contacting Foxglove	Flexible architecture. Supports on-prem, air-gapped, public and private cloud VPCs, and hybrid environments.
Extensibility framework	Custom panels, scripts, converters, SDK, API, webhooks, and an MCP server	Full-Stack JS/Python plugins. Custom components for front-end panels and Python operators for backend compute.

“FiftyOne has helped us speed up investigations by 3x. For example, if we see a wrong suction cup grasping an item, we can quickly visualize the issue across all data sources and identify what went wrong.”
Dimitry Pechyoni
Senior Principal Machine Learning Engineer at Berkshire Grey

Read the case study

"With FiftyOne, we were able to cluster 45,000 images from production for an analysis, and instantly found anomalies: products stacked on top of each other, unexpected artifacts, etc. This would've taken days to detect manually. The value that FiftyOne provides is priceless!" — Principal AI Engineer/Data Scientist
Principal AI Engineer, Fortune 500 Health Tech

“FiftyOne has become an important element of our dataset and model development pipeline. Since incorporating the tool into our workflows, we gained a better overview over what data is the best to be used to train and evaluate our models. As a result, we can deliver continuous dataset optimizations into production faster, as our model training pipelines read data directly from FiftyOne.”
Dmytro Prylipko
ML Engineer, EvoLogics

“What really stands out about FiftyOne is the flexibility. The plugin framework lets us customize our workflows based on our unique needs, and the mature SDK lets us consolidate more of our pipeline into one tool, avoiding the cost of stitching together multiple systems. FiftyOne integrates directly into our production pipeline to drive 80% reductions in workplace incidents.”
Patrick Rowsome
Head of Computer Vision Operations, Protex AI

Read the case study

"Some of our dips in model performance were because of edge cases. With FiftyOne, we were able to catch them the same day we analyzed model performance, something that would have taken a week otherwise."
Terrance Whitehurst
ML Researcher, FloVision

Read the case study

Break down data silos with a flywheel you won't outgrow

Point solutions might get your first prototype off the ground, but ad-hoc stacks turn into fragmented data silos as your data scales.

In the State of Physical AI report, 97% of teams struggle to iterate on datasets, but exceptional teams spend nearly 3x more time on data work. These teams don't just collect data. They build a structured, centralized infrastructure to continuously understand, curate, and evaluate it.

Voxel51 replaces fragmented tooling with unified data infrastructure built to scale alongside your robotics development. It reads MCAP and Hugging Face LeRobot datasets natively, so robot logs and teleoperation demonstrations live in one place.

Turn robot data into better models

Join leading robotics and physical AI teams using Voxel51 to query, curate, annotate, and evaluate multimodal sensor data at scale. Turn recorded episodes into the datasets that train better VLA and perception models.

Terms of ServicePrivacy Policy