The short answer
Why use Count Objects?
Counting people in a crowd shot or cars in a car park by hand means tapping and keeping track. The model marks each one it finds, so you only check the boxes instead of counting from scratch.
Detection confidence starts at 60% and runs from 10% to 95%. Each object the model is at least that sure of gets a red box and a number, and the status line says how many were found. Lower the value if obvious items are missed; raise it if shadows or background clutter get boxed.
The count mixes every category. A street photo with eight people and three cars reads 11, and the boxes carry numbers, not labels. Crowded or overlapping items are often merged into one box, and anything very small in the frame is easy to miss. Crop tightly to the area you care about with Crop Image before counting.
If you want to hide the faces you find rather than count them, Blur Face detects faces and blurs each area. To count pixels of a certain kind instead of objects, Pixel Counter reports how many pixels in an image are not transparent. The detection model behind Count Objects is loaded into your browser and runs there, so the photo stays on your device.
Where Count Objects fits
Event headcount
Get a rough count of people in a room photo taken from the back.
Car parks
Count parked cars in a photo of a small lot to check how full it is.
Table settings
Count the cups, bottles or chairs laid out before guests arrive.
Start to finish
How to use Count Objects
- 01Tap Choose image and pick a clear photo where the items do not overlap much.
- 02Leave Detection confidence at 60%, or lower it if items are missed; turn on Include manual correction box to add one.
- 03Tap Apply changes, check the numbered red boxes and the total, then tap Download result.
What to expect from Count Objects
| Check | What it means here |
|---|---|
| Supported input | JPEG, PNG, WebP, GIF, BMP, SVG and browser-supported AVIF images |
| Main controls | Confidence threshold and editable boxes |
| Output | A marked or blurred image and the detection details |
| Working limit | 50 MB and 50 megapixels per image; the model needs more memory than a simple filter |
Count Objects tips
- The detector stops at 30 objects; split bigger scenes with Image Splitter.
- Below about 30% confidence, shadows and patterns start getting boxed.
- Boxes show numbers only, so check the total includes only the type you meant.
- Crop tightly first; tiny items in a wide shot are often missed.
Count Objects FAQ
Something missing? Ask us.
What kinds of objects can it count?
Why is the count capped?
Can I add an object it missed?
Does Count Objects upload my file?
Is Count Objects free to use?
Why did Count Objects fail or look wrong?
Related tools
More in Measurement.