Skip to content
Metrica Docs

Glossary

Plain-language definitions of the key terms in the Metrica system — cameras, networking, computer vision, business logic, backend, and system flow. If you understand this glossary, you understand most of the system design.

Video & camera terms

Section titled “Video & camera terms”
Term Definition
CCTV (Closed-Circuit Television) A traditional security camera system for monitoring places like stores or malls. Not publicly accessible; used for surveillance and recording; usually connected to DVR or NVR systems.
IP camera (Internet Protocol Camera) A modern camera that sends video over a network. Has an IP address like a computer, streams video digitally, and can be accessed via protocols like RTSP.
RTSP (Real-Time Streaming Protocol) A protocol used to stream live video from IP cameras. Works over the local network, or over the internet if exposed.
NVR (Network Video Recorder) A device that receives, stores, and manages video from IP cameras. A central hub for multiple cameras that can provide RTSP streams; common in stores and malls.
DVR (Digital Video Recorder) The older version of an NVR, used for analog cameras. Less flexible than an NVR.

Example RTSP URL:

Terminal window
rtsp://username:password@192.168.1.10:554/stream

Networking terms

Section titled “Networking terms”
Term Definition
IP address A unique address for a device on a network (e.g. 192.168.1.20). Used to identify cameras or NVRs. Local IPs are not accessible from the internet.
Port A communication endpoint on a device (e.g. 554, the RTSP default). Used to access services such as video streaming.
LAN (Local Area Network) A private network inside a building. Cameras usually live here.
VPN (Virtual Private Network) A secure connection that lets you access a private network remotely — makes you “appear inside” the store network. Used for remote camera access.
Port forwarding A router configuration that exposes a local device to the internet. Rarely used in stores; a security risk if misconfigured.

Computer vision terms

Section titled “Computer vision terms”
Term Definition
Computer Vision (CV) The field of AI that lets machines understand images and video. Metrica uses CV to detect people, track movement, and generate events.
Frame A single image extracted from a video. Video is a sequence of frames; the AI processes frames, not raw video.
FPS (Frames Per Second) Number of frames processed per second. 30 FPS is smooth video; ~10 FPS is enough for analytics.
Object detection Identifying objects in an image — e.g. detect “person” and draw a bounding box around them.
YOLO (You Only Look Once) A real-time object detection model. Detects objects in images instantly and outputs bounding boxes plus confidence scores.
Bounding box A rectangle drawn around a detected object, showing where a person is in the frame.
Confidence score A number from 0–1 showing how sure the model is (0.9 = very sure, 0.3 = uncertain).
Object tracking Following the same object across multiple frames.
ByteTrack A tracking algorithm used to assign consistent IDs to people across frames (e.g. a person stays ID 5).
Track ID A temporary identifier assigned to a detected object. Not a real identity — used for tracking movement only.
Occlusion When objects block each other in the camera view (one person hides another). Causes tracking errors and lost detections.
Re-identification (ReID) An advanced method to recognize a person after they disappear. Not required for MVP; used in advanced systems.

Example YOLO detection output:

{
  "class": "person",
  "bbox": [x1, y1, x2, y2],
  "confidence": 0.92
}

Business logic terms

Section titled “Business logic terms”
Term Definition
Virtual line A line drawn in the camera view, used to detect entry and exit.
IN event Triggered when a person enters the store.
OUT event Triggered when a person leaves the store.
Session A complete visit of a person: one IN followed by one OUT equals one session.
Dwell time Time a customer spends inside the store: dwell_time = exit_time - entry_time.
KPI (Key Performance Indicator) A business metric — e.g. number of visitors, average stay time, peak hours.

Backend terms

Section titled “Backend terms”
Term Definition
API (Application Programming Interface) A system that lets the frontend and backend communicate.
FastAPI The Python framework used to build Metrica’s APIs quickly.
WebSocket A real-time communication channel used for live dashboards and instant updates.
PostgreSQL The relational database used to store structured data — events, sessions, and metrics.

System flow terms

Section titled “System flow terms”
Term Definition
Pipeline The full process from camera input to dashboard output: Camera → Detection → Tracking → Events → Database → Dashboard.
Latency The delay between a real-world action and the system’s response. Lower latency means a better real-time system.
Edge computing Processing data near the source (inside the store). Reduces network dependency and speeds up processing.
Section titled “Related pages”