Methodology

How CrossingWise estimates border waiting times

CrossingWise estimates border waiting times using recent queue observations, vehicle counts, queue movement, estimated processing rates, camera freshness, and historical patterns. These values are independent estimates and should not be treated as official border information.

Camera-derived estimate

Waiting times are independent estimates based on vehicles visible in a public camera image and estimated border throughput. The queue may extend beyond the visible frame.

Vehicle counts

Come from a YOLO object-detection model run on a still frame captured from each crossing's public traffic webcam every minute. It's a rough estimate of what's visible in frame, not an official measurement, and can miss vehicles that are out of shot, occluded, or too distant to detect reliably.

Estimated wait

Queue length ÷ a vehicles-processed-per-hour figure for that crossing. That figure starts as a guess, but is continuously recalibrated in the background against borderalarm.com's reported wait times once enough of them line up with a camera reading — until then, or for crossings borderalarm.com doesn't cover, it stays a guess.

A “≥” prefix

Means the visible queue has stayed at the camera's practical viewing limit for several samples in a row, so the real queue (and wait) likely extends further than what's shown.

Community-reported wait

A recently reported wait time from an external, crowd-sourced source — shown for context, never blended into the camera-derived estimate above.

Reported wait pills

Entry/exit figures are crowd-sourced from borderalarm.com, refreshed at most every 5 minutes.

Webcam footage

Sourced from uzivokamere.com.

Confidence, trend, and forecast

Confidence level: Reflects camera freshness, visibility, and detection quality — not an official measurement.
Queue trend: Compares recent observations to flag whether a crossing's queue looks to be increasing, decreasing, or holding steady.
Historical pattern: “Typical queue by hour” is the median observed queue for that local hour of day, computed from past observations at that crossing.
Forecast: A short-term prediction from this app's own baseline model, not a current measurement — treat it as a rough guide, especially further out. It's withheld entirely when the camera or current estimate isn't trustworthy enough to build on.
Data freshness: Shown per crossing (e.g. “Live”, “Updated N min ago”, “Stale”, “Camera offline”) based on how old the last successful camera read is.

Kinds of data shown

This app shows four distinct kinds of information, never blended into one number: camera-derived estimates (this app's own vehicle counts and wait ranges), community reports (unverified, user-submitted), external reported data (crowd-sourced figures from borderalarm.com), and forecasts (predictions from this app's own baseline model).