Get startedScore Studio · documentation
Start here. Know what happens next.
Analyze an image or clip, review labels, train a model, and test it before deployment. Start with the product guide if you are new to Score Studio. Each guide explains inputs, actions, results, and availability limits.
Product documentation
Start
Local
Run Score Studio locally
Run the Apache 2.0 software on infrastructure you operate. You supply storage and inference; help comes through community docs and discussions.
Product map
Understand Score Studio
Start with an image, a clip, or existing data. Learn what is saved, what runs automatically, and what needs your approval.
Product documentation
Workflows
Design and orchestration
Build a production workflow
Compose the complete vision lifecycle instead of treating model training as the final destination.
Durable execution
Run and control workloads
Follow annotation, model analysis, evaluation, and workflow jobs with saved progress and results.
Preview and tracked runs
Choose the right execution mode
Use Preview for transient inspection and a tracked workflow for saved labels, artifacts, approvals, deliveries, and receipts.
Product documentation
Data
Versioned evidence
Build and inspect datasets
Create specialist or multimodal datasets around the outcome the system must understand.
Label and review
Use Annotation studio
Draw or correct object labels on images and sampled frames, then explicitly confirm their review.
Portability
Import and export annotations
Move labels and multimodal training evidence between Score and external systems.
Statistics and quality
Diagnose dataset health
Use coverage, balance, dimensions, spatial plots, and split statistics to find data risk before training.
Asset operations
Manage dataset assets
Filter, archive, restore, and permanently remove versioned media and annotations safely.
Product documentation
Models
Artifacts and foundations
Choose and inspect models
Choose a ready specialist model or an off-the-shelf vision-language foundation, then inspect the exact runtime and version.
Availability limit
Understand VLM training availability
Multimodal dataset schemas do not imply an executable vision-language model trainer. General VLM fine-tuning is not an implemented training path in this release.
Local training
Train a model to outline objects
Segmentation learns object shapes. Training your own model is separate from using the built-in segmentation service.
Experimental · local CPU
Optimize vision-model weights with rewards
This experimental path uses reviewed annotations to reward sampled model outputs and update the vision model itself.
When training stops
Recover a failed training attempt
Find the cause and keep the exact data selection instead of starting another run blindly.
Optimization
Train and fine-tune models
Train from reviewed boxes or object outlines. Detection candidate search uses a configured engine; segmentation and experimental reward learning run locally.
Quality evidence
Evaluate a model release
Measure one executable model version on a declared benchmark or custom held-out dataset with frozen thresholds, metrics, uncertainty, and sample evidence.
Frontier model work
Commission The Vision Lab
Ask Score to build a frontier or unusually complex visual model when existing foundations are insufficient.
Interactive validation
Test a model interactively
Run representative visual inputs through a selected model before committing it to a workflow or release.
Product documentation
Operate
Release and runtime
Deploy and operate models
Release models to devices, providers, managed APIs, or outcome-based workloads.
Continuous assurance
Run conformity and monitoring
Check live behavior against a declared production contract and investigate drift or policy failures.
Serving
Create and manage deployments
Serve an evaluated model or workflow through managed, provider-backed, local, or edge infrastructure.
On-device runtime
Register and operate edge devices
Connect owned hardware for local inference, private processing, and device-backed workloads.
Global feedback
Track and control tasks
Follow every long-running operation from one persistent workspace task center.
Product documentation
Developers
External systems
Connect providers and integrations
Connect infrastructure once. Score verifies it and publishes its exact capabilities to every compatible product surface.
Automation
Build with the SDK and REST API
Automate the same data-to-production lifecycle used in the Score Studio interface.
MCP access
Connect agents with controlled execution
Connect an MCP-compatible agent to workspace-scoped datasets, models, evaluations, and deployments.
Live OpenAPI
REST API reference
Inspect schemas and execute authenticated API operations against the configured Score Studio API.
Product documentation
Workspace
Product guides are versioned with Score Studio so behavior, safeguards, and documentation ship together.