How to count people in a photo: 3 methods that actually work

Updated July 2026 · 5 minute read

Whether you're reporting event attendance, settling an argument about crowd size, or keeping records for your organization, you eventually face the same problem: a photo full of people and a need for a real number. Here are the three methods that work, from free-and-slow to instant.

Method 1: The grid count (accurate, painfully slow)

Open the photo in any editor, overlay a grid (or just zoom in section by section), count each cell, and add them up. Marking counted heads with dots as you go prevents double-counting.

Method 2: The Jacobs density method (fast, rough)

The classic crowd-estimation technique from journalism: estimate the area the crowd occupies, assign a density (light crowd ≈ 1 person per 10 sq ft, dense ≈ 1 per 4.5 sq ft, packed ≈ 1 per 2.5 sq ft), and multiply.

Method 3: AI crowd counting (seconds, verifiable)

Modern crowd-counting models do what the grid method does — find every person in the image — but in seconds instead of hours. Here's a real example: packed gym bleachers, photographed from the scorer's table, run through Crowd Counter's model.

Packed gym bleachers at a college basketball game, before counting The same bleachers with a detection dot on every person found by the AI model
Left: the photo. Right: the model's detection map — every dot is one person it found. Result: 734 people, counted in about twelve seconds.

The detection map is the part that matters. A number alone asks you to trust the software; a map of every counted person lets you verify it yourself — and if you disagree, you correct the count by hand and keep your number.

Count your next crowd from one photo — your first count is free.

 Download Crowd Counter

Tips for photos that count well

  1. Shoot from elevation. A balcony, press box, or raised corner reduces the overlap that hides people behind each other.
  2. Get the whole crowd in frame. Two photos of two sections beat one photo that crops people out.
  3. Steady and wide. Blur is the main thing that hurts machine counts — brace the phone and avoid zooming.
  4. Count at the same moment every time. If you're tracking attendance across events, consistency of timing matters more than any single count's precision.

One-off curiosity counts are fine — but if you count the same event every week or every game, the real value is the record: every count filed to its season with the photo behind it, trends over time, and a report you can export when someone asks. That's exactly what Crowd Counter was built for.