Our Technology

From Edge Vision to City‑Scale Intelligence

T²i-Flow fuses edge CV, streaming analytics, and predictive optimization into an end‑to‑end control loop.

Pipeline Overview

Our vision stack spans the full lifecycle: acquisition, representation, feature extraction, classification, and post‑processing. Each stage is optimized for low‑latency operation at the edge with central orchestration.

Computer vision pipeline diagram from image acquisition through post‑processing — Acquisition → Representation → Feature Extraction → Classification → Post‑processing

Core Techniques

We combine classic image processing with deep learning. Below is a map of common operators we deploy where appropriate to improve SNR before inference or to post‑process results.

Mind‑map of image processing techniques: compression, filtering, morphology, color space conversion, edge detection, histogram equalization, resizing/transform — Classical operators remain invaluable for stability and speed.

Platform pipeline

Capture

Edge CV, ANPR, sensors

Ingest

Streams, QoS, schemas

Analyze

Forecasts, detection

Optimize

OR, simulation

Actuate

Signals, tolling, APIs

Multi‑view vehicle detection, tracking, re‑ID
ANPR/OCR with confidence scoring
Privacy filters (face/plate masking variants)

Event schema (example)

{
  "event": "vehicle_detected",
  "ts": "2025-08-10T12:34:56Z",
  "site_id": "I5-NB-23A",
  "lane": 3,
  "bbox": [0.12, 0.44, 0.21, 0.18],
  "speed_kph": 92.3,
  "plate": { "text": "7XYZ123", "confidence": 0.997 },
  "classes": ["car"],
  "hash": "sha256:..."
}