Camera placement for accurate face recognition in gambling venues

Face recognition is only as good as the images it receives. Most accuracy problems in real venues come down to three things: where the camera is, how the face is lit, and how big the face appears in the frame. Get those right and performance improves dramatically.

This guide draws on our support documentation and experience of real installations.

1. Choose the right camera

• Resolution: 1080p minimum. Face detection and recognition depend on image quality. 720p cameras will work, but accuracy may be lower than you need.
• Lens matters. A lens that is too wide makes faces small; a narrower field of view on the entrance gives more detail. Contact us for lens advice for your layout.
• USB or IP. ClientScan works with standard USB webcams and RTSP / IP cameras, so you can often use existing hardware.

2. Put it where faces are predictable

The best position is where people walk towards the camera and look roughly straight at it — usually the entrance or an internal doorway.

• Mount at approximately head height, angled only slightly downward.
• Avoid ceiling-mounted "bird's-eye" views — they capture the top of the head, not the face.
• Make sure the field of vision isn't too wide; you want faces, not the whole room.

3. Get the lighting right

Lighting is the most common cause of poor results:

• Avoid backlight. A bright doorway or window behind the customer turns faces into silhouettes.
• Avoid side lighting that leaves half the face in shadow.
• Aim for stable, even illumination that doesn't change dramatically through the day.

4. Make faces big enough

As a rule of thumb, the face should take up at least a quarter of the height of the video frame shown in the application. Smaller faces may still be recognised, but with lower confidence. Use zoom or reposition the camera if faces appear small.

Alignment matters too: accuracy can fall by around 30% if a face is misaligned by more than about 10% — another reason to position the camera where people naturally look towards it.

5. Give the system time

The longer a face is visible, the more chances the system has to make a confident match. A short approach corridor or a door that customers walk straight through works better than a camera that catches them for a fraction of a second.

6. Keep the database healthy

• Review stored images regularly.
• Delete poor-quality images.
• Remove duplicates, or group them under the same person.
• When you upgrade cameras, capturing new images improves accuracy further.

Quick checklist

• 1080p camera or better
• Mounted at head height at the entrance
• No bright backlight or strong side lighting
• Faces fill at least a quarter of the frame height
• Customers walk towards the camera for a second or two
• Database reviewed regularly

Need help with a specific site? Our support page covers common questions, or get in touch.