VPEP: Video-to-Protocol Extraction Pipeline
Upload a laboratory experiment video and its corresponding Standard Operating Procedure (SOP). Our autonomous VLM pipeline will extract actions, verify compliance, and generate a bilingual interactive dashboard.
Pipeline Architecture (6 Stages)
Key Capabilities
Automated Protocol Audit
VLM-powered analysis extracts every hand movement, reagent, and equipment interaction with sub-second precision.
Bilingual EN/JA Reports
Generates comprehensive reports in both English and Japanese, including executive summaries and compliance matrices.
Inventory Tracking
Automatic material consumption ledger tracks every reagent volume and consumable used during the experiment.
VLM-Powered Keyframe Analysis
Extracts and annotates critical keyframes with bounding boxes, action labels, and hand tracking overlays.
Technical Features
GCS Direct Upload
Videos upload directly to Google Cloud Storage via signed URLs, bypassing Cloud Run's 32 MB limit.
6-Stage Autonomous Pipeline
Fully autonomous execution from video ingestion to interactive dashboard generation — no manual intervention.
Interactive HTML Dashboard
Embeddable single-file HTML report with timeline navigation, keyframe gallery, and protocol compliance heatmap.
Intelligent Compression
FFprobe-based detection automatically skips re-compression for already-optimized videos, saving compute time.