GreenOcean AI
Computer-vision traffic analytics, billed by usage.
A multi-tenant computer-vision platform for traffic studies. Clients upload camera-view videos and get accurate, per-site vehicle-count reports — usage metered per minute and billed through Stripe.
What it does
End-to-end ML pipeline for ingesting camera feeds, preprocessing video, and deploying models
Automated training workflows with data versioning, augmentation, and validation loops
Computer vision inference for detection, tracking, classification, and vehicle counting
Custom-trained YOLO detector with BoT-SORT multi-object tracking
Turning-movement counts, ADT, vehicle classification, and speed studies
SAM-based auto-labeling training loop (SAM 2 in production, SAM 3 in evaluation)
Token-metered usage tracked per user and per organization
Stripe billing — flat PRO tiers and Enterprise rate cards
Live count-over-time dashboards, peak-hour analysis, and exportable Excel reports
Role-based org accounts with admin controls and project/site management
Google Maps project & camera-view placement
Technologies27
PythonPyTorchYOLO (Ultralytics)BoT-SORT (boxmot)SAM 2 / SAM 3OpenCVShapely / NumPy / SciPyONNX RuntimeHugging Face Transformers.jsRoboflowffmpegApache AirflowAWS ECS (GPU g4dn / Fargate)AWS LambdaAmazon S3AWS RekognitionAmazon CognitoTerraformGoogle MapsNext.jsReactTypeScriptTailwind / RadixPrismaPostgreSQL (RDS)DynamoDBStripe
Want something like GreenOcean AI?
We design and build custom software end to end.