TSING-LabelOps
Qingbiao Hub — a labeling & QA platform for embodied-AI data production
Connecting the TaskHall, supplier uploads, raw-video review, task dispatch, clip cutting, labeling workbench, QA review and data export — turning first-person capture into stable, trainable data assets.
The middle layer between TaskHall and data products
TSING-LABELOPS — MIDDLE LAYERTaskHall publishes collection demand to collectors; Qingbiao Hub turns what comes back into a controlled production flow — raw-video review, supplier management, labeling dispatch, clipping, QA inspection and export tracking.
A configurable, auditable, measurable production flow
Production Dashboard
Filter by Provider / Scenario / Environment and watch key metrics and progress curves in real time — ingestion, dispatch, labeling, Clips, exports.
Raw Video Review
Audit collected data for task consistency, privacy compliance, visual quality and labelability — with Approved / Returned / Needs Changes / Discarded states.
Batch & Workforce Scheduling
Dispatch reviewed videos to annotators, suppliers or QC staff in batches, with task-status tracking and annotation-capacity management.
Timeline Clip Workbench
Mark one or more Clips on the video timeline; each Clip binds Task / Subtask / Quality and scene environment into a trainable segment.
Inspection & Re-review
Sample-check, re-review, return and correct labeled tasks, recording annotator quality performance for stable, usable output.
Data Product Interface
Export approved clips, labels, metadata and quality reports to the data product line — feeding HM, HOI, DexSkill and Real2Sim products.
Multi-role collaboration and a closed quality loop
Admins, suppliers, annotators, reviewers and QC staff each touch only their own step, yet every action is uniformly traced — a closed evidence chain from collection to data product.
Role Isolation
Admins, suppliers, annotators, reviewers and QC staff each own their scope, with isolated permissions.
Full Traceability
Ready → Dispatched → Waiting Review → Approved → Needs Changes → Abandoned → Exported — status trackable end to end.
Quality Binding
Every Clip carries a Quality grade — used for training-data selection, payment settlement and annotator evaluation.
Auditable History
Retains Provider / Scenario / Environment, batches, annotators, review records and full modification history.
Turn captured data into trainable assets, reliably
Book a TSING-LabelOps walkthrough and see the controlled flow from raw video to data products.