My tools

Local AI · Counting

AI Object Counter: count people, cars, bottles and more in a photo

How many people came, how many cars are parked, how many bottles are on the shelf? Drop a photo and AI on your device boxes and numbers every object. If it misses or misreads something, one click fixes it.

Local AI · no uploads

Privacy: Images are processed only in your browser and never uploaded to a server; nothing is kept after you close the page.

Drop images here, click to choose, or paste with Ctrl+V

YOUR NEXT STEP

Finish your materials, then connect the rest of your work

Use image and document features under their stated free limits. Create a company account for website monitoring, business inquiries and team notifications, or discuss your workflow with us.

New companies receive a 14-day Pro trial, with no card required. Trials exclude the paid monthly credit grant.

Already have an account? Open your workspace →

How to use the AI Object Counter

  1. Drop or take a photo.
  2. AI boxes objects and counts them by class; tick “Find small objects” for crowds or tiny items.
  3. Pick the class to count and move the threshold to drop unreliable boxes.
  4. Click empty space to add a manual mark; click a box or mark to exclude it. Export the annotated image or CSV.

What can it count?

The 80 everyday COCO classes: people, bicycles, cars, motorbikes, buses, trucks, boats, traffic lights, benches, birds, cats, dogs, horses, cows, sheep, backpacks, umbrellas, suitcases, bottles, cups, bowls, cutlery, bananas, apples, oranges, carrots, pizza, cake, chairs, sofas, potted plants, TVs, laptops, phones, books, clocks and more. For things outside the list, such as pills, screws, seeds or rebar, use manual counting.

Getting accurate counts

Large, sharp, unobstructed objects count best. Dense crowds and small distant items are easy to miss. “Find small objects” splits the photo into overlapping tiles, detects each and merges duplicates; it is slower but finds more. A lower threshold finds more objects and more false hits; a higher one does the opposite.

Manual fixes and export

After the automatic count, click where the AI missed something to add a numbered purple mark. Click a box to exclude it (dashed) and click again to restore it. Total = valid boxes + manual marks. The exported image is numbered and the CSV lists every box with class, confidence and coordinates for spreadsheets or records.

Privacy

Counting happens in your browser and photos are not uploaded. If you will publish a crowd photo, run it through Face Blur afterwards.

Frequently asked questions

Are photos uploaded?

No. The model downloads from this site and runs in your browser.

How many objects can it count?

Each pass keeps up to about 300 boxes, and tiled detection finds more small ones. For extremely dense scenes, such as thousands of people, treat the result as an estimate.

Why does one object have two boxes?

Tiled detection can see objects on tile edges twice. Overlapping boxes of the same class are merged automatically; click any remaining duplicate to exclude it.

Can it count video or webcam frames?

Photos only for now.

It can’t detect what I need to count.

Use manual counting: click each object and the tool numbers and totals them.

Can I use the result as an official count?

AI counting is an aid; have a person verify official figures.

AI Object Counter practical guide and common questions

What does an AI Object Counter do?

The 80 everyday COCO classes: people, bicycles, cars, motorbikes, buses, trucks, boats, traffic lights, benches, birds, cats, dogs, horses, cows, sheep, backpacks, umbrellas, suitcases, bottles, cups, bowls, cutlery, bananas, apples, oranges, carrots, pizza, cake, chairs, sofas, potted plants, TVs, laptops, phones, books, clocks and more. For things outside the list, such as pills, screws, seeds or rebar, use manual counting.

Who is this for, and when is it useful?

People who need to count people, vehicles, goods or parts in a photo without tallying them by hand.

How do I use this AI Object Counter?

Drop or take a photo. AI boxes objects and counts them by class; tick “Find small objects” for crowds or tiny items. Pick the class to count and move the threshold to drop unreliable boxes. Click empty space to add a manual mark; click a box or mark to exclude it. Export the annotated image or CSV.

How should I interpret the results?

Large, sharp, unobstructed objects count best. Dense crowds and small distant items are easy to miss. “Find small objects” splits the photo into overlapping tiles, detects each and merges duplicates; it is slower but finds more. A lower threshold finds more objects and more false hits; a higher one does the opposite.

Where is the work performed, and which services receive data?

AI Object Counter: The recognition models are downloaded from this site and run in the browser; photos and results are processed only on your device and never uploaded, and any saved log stays in this browser.