We use cookies.This website uses essential cookies to operate core features. With your consent, we also use analytics cookies to understand traffic and improve the service. For more details, see our .
We use cookies.This website uses essential cookies to operate core features. With your consent, we also use analytics cookies to understand traffic and improve the service. For more details, see our .
Was this tool helpful to use?
Your feedback helps us make it better
Use AI to count the number of people in an image. Supports JPG/PNG uploads and image resizing for accurate counting results.
Choose filesDrag and drop or click to upload a photo (PNG / JPG / WEBP)
Image (Supported formats: .jpeg, .jpg, .png, .webp)
Max 1 MB
Face detection runs locally in your browser using MediaPipe, so your image is never uploaded. The first run downloads a model of about 10 MB. The model is designed for single-face detection at short range, so it is less accurate for distant crowds. Close-up and mid-range group photos work best.
Overview
Understand what the tool solves, how it works, and the boundaries of its data.
This tool accepts one still image and returns the number of faces it detects, with a numbered rectangle drawn around each detection. It is a face counter, not a general headcount system: people whose faces are turned away, hidden, too small, or outside the frame may not be counted, while faces on posters or screens can be detected as faces too.
The underlying face-detection task locates faces in an image and returns bounding boxes. The result is a model output, not a verified count of people in a room, crowd, or event. The current upload accepts PNG, JPG/JPEG, or WebP files up to 1 MB, one image at a time.
Technical reference
Guide
Follow the workflow and verify inputs and outputs with practical examples.
Select a PNG, JPG/JPEG, or WebP image no larger than 1 MB. Upload one file for each run.
Start detection after the preview appears. Wait for the count and overlay to finish; the tool marks detected faces with green boxes and sequence numbers.
Inspect the full image for missed faces, repeated detections, faces in background pictures, and faces that should not be part of your intended count. Choose another image to clear the current result and start again.
For a more useful visual check, use a sharp image with faces large enough to inspect and avoid cropping people at the frame edges.
Use cases
See how the tool fits into real work and everyday tasks.
A teacher can use the boxes to make a quick face-by-face check of a class picture, then compare it manually with the expected group rather than treating the model count as attendance.
A photographer can use the overlay as a visual prompt to notice a face that may be hidden behind another person or placed near the edge of the frame.
An organizer can explore how many faces are visible in a single still image. It does not count people across multiple frames or establish venue occupancy.
Q&A
Find concise answers to common questions and confusing cases.
No. It counts detected faces. A person whose face is hidden, turned away, very small, blurred, or cut off may be missed. Review the image and boxes yourself.
It can be. The detector locates faces in the image; it does not know whether a face belongs to a live person or is printed or displayed in the scene.
No. A single still and face detections cannot verify attendance or the number of people physically present. Use an appropriate, reviewed counting process for those decisions.
Notes
Review scope, result limitations, and important precautions before use.
Face detection can miss or misidentify faces because of image quality, lighting, pose, occlusion, distance, or background content. The result is not a safety, security, access-control, legal, or occupancy measurement. Do not use it as the sole basis for attendance records, crowd limits, or decisions about identifiable people. Obtain permission to use images where appropriate and follow the rules that apply to your setting.
Related
Discover related tools, collections, and available API capabilities.