Dahua DSS Pro Facial Recognition Setup Guide (2026) — Complete Step-by-Step

Facial recognition in DSS Professional 8.7 is one of the platform’s most powerful features — and one of the most frequently misconfigured. Many deployments go live with FR technically enabled but effectively non-functional: wrong camera placement, threshold set too low, person library not populated, or the alert rule pointing at the wrong channel. This guide walks through every step from license verification to live watchlist alerting, so you get FR working correctly the first time.

---

What You Need Before You Start

Before touching any configuration in DSS Pro, confirm these three prerequisites are in place:

1. FR License

Facial recognition in DSS Pro is a licensed feature. The base DSS Pro license (DSS8PRVB) includes 2 face recognition channels. Each additional FR channel requires a separate add-on license. The number of FR channel licenses you need equals the number of cameras that will perform FR processing — not the total number of cameras in the system.

To verify your current FR license allocation: open the DSS Pro client → Maintenance → License Management → check the “Face Recognition” row. If the count shows 0, FR will not activate regardless of camera or configuration.

2. Compatible Camera Hardware

Not all Dahua cameras support facial recognition. DSS Pro supports FR in two distinct modes:

Mode	Where processing happens	Camera requirement	Server load
Edge FR	Camera (on-device AI chip)	Dahua WizMind / AI series with built-in FR	Low
Server-side FR	DSS Pro server	Any camera with clear face capture (720p minimum)	High (GPU recommended)

Edge FR (recommended) uses cameras with an embedded AI processor — primarily the Dahua WizMind line (e.g., IPC-HDW8849H-EHL, IPC-HFW8849S-Z4-LED, IPC-HFW8869S-Z4-LED). These cameras perform face detection and feature extraction on the device and send only the metadata and face snapshot to DSS Pro for matching. This dramatically reduces server load and allows real-time matching even on large person databases.

Server-side FR uses DSS Pro’s server to extract facial features from the video stream. It works with a wider range of cameras but requires significantly more server CPU/GPU resources. For deployments beyond 4–5 server-side FR channels, a dedicated GPU (NVIDIA recommended) is strongly advised.

3. Server Hardware

For edge FR (recommended): standard DSS Pro server specs are sufficient. For server-side FR at scale:

• Up to 4 channels: modern 8-core CPU, 32 GB RAM
• 5–16 channels: NVIDIA GPU with 8+ GB VRAM (e.g., RTX 3080, Quadro RTX 4000)
• 16+ channels: dedicated GPU server or multiple servers with load distribution

---

💬 Need a license or have questions? → Message us on Telegram — free consultation, usually reply within a few hours.

Step 1 — Camera Placement for Optimal FR Accuracy

FR accuracy is more dependent on camera placement than on software configuration. Poor placement is the single most common reason FR deployments fail to meet accuracy expectations. Dahua’s validated requirements:

Parameter	Minimum	Recommended
Face pixel width in frame	80 px	120–200 px
Face angle (horizontal)	±30°	±15° (near-frontal)
Face angle (vertical, tilt)	±20°	±10°
Camera mounting height	—	2.2 m – 3.0 m above floor
Camera-to-face distance	—	1.5 m – 5 m (varies by lens)
Illumination (lux)	100 lux	300+ lux, even lighting
Backlighting	Avoid	Supplemental front lighting if backlit

Practical placement tips:

• Mount at eye level or slightly above — never directly overhead. A camera mounted at 5 m height pointing down at a walking person will capture mostly the top of the head, not the face.
• Choke points work best: doorways, turnstiles, single-lane corridors. Wide open areas with multiple approach angles dramatically reduce FR hit rates.
• For entrance doors, position the camera so it captures the face as the person approaches, not after they have passed. A camera behind the door pointed at backs provides zero FR value.
• If supplemental lighting is needed, use even white light — colored or infrared-only illumination reduces FR accuracy on most Dahua AI cameras.

---

Step 2 — Add the FR Camera to DSS Pro

If the camera is not yet in DSS Pro, add it via Device Management → Add Device. Select the correct protocol — use the Private (Dahua) protocol for WizMind cameras, not ONVIF. ONVIF connections do not pass AI metadata and will prevent FR from functioning even if the camera supports it natively.

Once added, verify the camera appears under Device Management with a green status icon. If the status is yellow or red, resolve the connection issue before proceeding — FR configuration on an offline device will not save correctly.

---

Step 3 — Enable Face Recognition on the Channel

1. In the DSS Pro client, navigate to AI Config → Face Recognition → Channel Config.
2. Click Add and select the camera channel you want to assign to FR.
3. Set the FR Mode:
   • Local (Edge) — for WizMind cameras with on-device FR processing
   • Central (Server) — for non-AI cameras using server-side processing
4. Set the Detection Region — draw a zone within the camera frame where face detection is active. Restricting the zone to the choke point (e.g., just the doorway rectangle) prevents false detections from background activity.
5. Set the Minimum Face Size threshold (pixels). Set this to at least 80 px; 120 px recommended.
6. Enable Face Filter options:
   • Blur filter — reject blurry face captures (recommended: ON)
   • Occlusion filter — reject faces with masks/glasses below threshold (configure based on your environment)
   • Angle filter — reject faces beyond a set yaw/pitch angle (recommended: ON, ±30°)
7. Click Save.

After saving, the channel appears in the FR channel list with its assigned FR license slot consumed. Verify the license count drops by 1 under Maintenance → License Management to confirm the FR license is correctly applied to this channel.

---

Step 4 — Build the Person Library

The Person Library is the face database DSS Pro matches against. DSS Pro 8.7 supports up to 100,000 persons in the database. This step is where most deployments underinvest — a well-populated, well-organized library is what separates a useful FR system from a false-alarm generator.

4a. Create Person Groups

1. Navigate to Person Management → Person Library.
2. Click Add Group and create logical groups for your deployment. Examples:
   • Staff — Building A
   • Staff — Management
   • VIP / Authorized Visitors
   • Watchlist — Banned Persons
   • Watchlist — Persons of Interest
3. Assign each group a Group Type: Allowlist, Blocklist, or Stranger. This type determines how DSS Pro handles a match — allowlist matches can be logged silently; blocklist matches trigger high-priority alarms.

4b. Add Persons and Enroll Faces

Persons can be added individually or in bulk:

Individual enrollment:

1. Inside a group, click Add Person.
2. Fill in the required fields: Name (mandatory), ID number, department, notes (optional).
3. Under Face, click Upload Photo. Upload a clear, front-facing headshot. DSS Pro will validate the face quality — if it rejects the image, the face is too small, blurry, angled, or contains multiple faces. Crop tightly to just the face and re-upload.
4. You can add multiple face photos per person (up to 5 recommended). Multiple photos from slightly different angles, lighting conditions, and with/without glasses significantly improve recognition accuracy in real-world conditions.
5. Click Save.

Bulk import:

1. Navigate to Person Management → Import.
2. Download the import template (Excel format) and fill it with person data: name, ID, group, photo file path.
3. Prepare photos as JPEG files named to match the person ID column.
4. Package photos and the completed template into a ZIP file and upload.
5. DSS Pro processes the import and reports success/failure per record. Records with rejected photos (quality failures) must be corrected individually.

Photo quality requirements for enrollment:

• JPEG format, minimum 100×100 px (300×300 px or larger recommended)
• Single face per image — group photos will be rejected
• Near-frontal angle — profile shots will fail quality validation
• Neutral background preferred; high-contrast busy backgrounds reduce feature extraction quality
• No severe occlusion — sunglasses, hats, or scarves covering the face will fail validation

---

Step 5 — Link Person Groups to FR Channels

Creating a Person Library does not automatically make DSS Pro match against it on any camera. You must explicitly link person groups to FR channels.

1. Navigate to AI Config → Face Recognition → Channel Config.
2. Select the FR channel you configured in Step 3.
3. Click Person Library tab.
4. Click Add and select the person groups this channel should match against.
5. Set the Similarity Threshold for each group:
   • 85% — recommended starting point for most deployments
   • 80% — increases recall (fewer missed matches) but increases false positives
   • 90%+ — high precision, but may miss genuine matches in suboptimal lighting
6. Click Save.

You can link multiple person groups to the same channel with different thresholds per group. A practical configuration: Blocklist group at 80% (you want fewer misses even at the cost of some false alarms), Staff Allowlist at 85% (balance), VIP group at 88% (precision matters for access control integration).

---

Step 6 — Configure Alert Rules for FR Matches

Without alert rules, DSS Pro logs FR matches silently in the database but does not notify operators. For watchlist-based deployments, alert rules are essential.

1. Navigate to Event Management → Event Config → AI Event.
2. Click Add, select event type Face Comparison Alarm.
3. Set the trigger conditions:
   • Channel: select the FR camera channel
   • Alarm Type: choose between Matched (person found in library) and Stranger (face detected but not in library)
   • Person Group: select which group triggers this rule (e.g., Blocklist only)
4. Set the Action:
   • Alarm on Monitor — pop-up on operator screen with face snapshot and library photo side-by-side
   • Send Email — email notification with snapshots
   • Trigger Output — trigger a relay (e.g., lock a door, sound an alarm)
   • Record Video — start recording on linked cameras when match occurs
   • Capture Snapshot — save snapshot to storage
5. Set a Schedule — define when this alert rule is active (24/7 or specific hours/days).
6. Click Save.

For security operations centers, configure the Blocklist rule to trigger Alarm on Monitor with Record Video on all cameras in the vicinity. This gives the operator the live context they need to respond immediately.

---

💬 Need a license or have questions? → Message us on Telegram — free consultation, usually reply within a few hours.

Step 7 — Test and Validate the FR Pipeline

Before going live, validate each stage of the pipeline:

Test 1 — Face Capture Verification

1. Navigate to Live View and open the FR camera.
2. Walk through the detection zone. In the FR panel (right side or AI panel), face snapshots should appear in real time.
3. If no snapshots appear: check camera connection protocol (must be Private/Dahua, not ONVIF), verify detection region covers the walkway, and check the minimum face size is not set too high.

Test 2 — Watchlist Match Test

1. Enroll a test person (yourself or a colleague) into the Person Library, placed in the Allowlist group.
2. Walk through the FR camera’s detection zone.
3. Navigate to AI Search → Face Comparison Search and query the last 5 minutes on that channel.
4. The result should show your face capture alongside your enrolled library photo, with a similarity score at or above the threshold.
5. If the match does not appear, lower the similarity threshold temporarily to 70% and re-test. If a match appears at 70%, your enrollment photo quality or real-world capture quality is marginal — improve one or both before raising the threshold back to 85%.

Test 3 — Alert Rule Trigger Test

1. Move the test person to the Blocklist group (or add a secondary test person to the Blocklist).
2. Walk through the detection zone.
3. Verify the alert pop-up appears on the operator’s monitor within a few seconds of detection.
4. Check that the side-by-side comparison (captured face vs. enrolled photo) displays correctly in the alarm popup.

---

Step 8 — Face Search and Retrieval

DSS Pro 8.7 stores all face capture events in the FR database. This enables powerful post-incident retrieval — you can search your recorded footage by face rather than scrubbing through hours of video.

Search by uploaded photo

1. Navigate to AI Search → Face Search.
2. Click Upload Photo and provide a reference face image (from any source — ID photo, social media, other camera, etc.).
3. Set the time range, channels, and minimum similarity threshold.
4. Click Search. Results show all face capture events with a similarity score above the threshold, sorted by score.
5. Click any result to jump directly to the video at that timestamp — no manual scrubbing required.

Search by person from library

1. Navigate to AI Search → Face Comparison Search.
2. Select the person from the Person Library.
3. Set time range and channels.
4. Results show all instances where this person was matched, including timestamp, channel, similarity score, and captured snapshot.

Stranger search

To find all persons in footage who are not in your library (useful for investigations):

1. Navigate to AI Search → Face Search.
2. Set Match Type to Stranger.
3. Set the time range and channels.
4. Results show all face capture events with no match in the Person Library.

---

Step 9 — Access Control Integration (Optional)

DSS Pro can integrate facial recognition with its access control module to create contactless face-based door entry. This requires:

• A Dahua face recognition terminal (e.g., Dahua DHI-ASI8213X-T1, DHI-ASI7213X-T1) added to DSS Pro as an access control device
• The terminal’s face library synchronized with the DSS Pro Person Library
• An access rule defined: persons in a specific group are granted access through a specific door

Configuration path: Access Control → Door Config → Face Recognition Access Rule. Select the door, link the person group, set the schedule, and enable sync. DSS Pro will push the enrolled face data from the Person Library to the terminal automatically.

This creates a unified system where the same person library used for camera-based watchlist matching is also used for door access — no separate enrollment per device.

---

Troubleshooting Common FR Problems

Symptom	Likely cause	Fix
No face snapshots appearing in live view	Camera added via ONVIF instead of Private protocol	Re-add camera using the Dahua Private protocol
FR channel shows “license not available”	FR license slots exhausted	Check license count in Maintenance → License Management; acquire additional FR channel licenses
Faces detected but no matches, even for enrolled persons	Person library not linked to FR channel	Step 5 — link person groups to the channel in Channel Config → Person Library
High false positive rate (too many wrong matches)	Similarity threshold too low	Raise threshold from 80% to 85–88% and re-evaluate
High false negative rate (enrolled persons not matching)	Enrollment photo quality poor, or camera angle/lighting suboptimal	Re-enroll with better-quality photos; review camera placement against criteria in Step 1
Alert rules firing but no popup on operator screen	Alert not assigned to a monitor group or the operator’s workstation	Check Event Management → Alarm Output → confirm the monitor target includes the operator’s client
Server CPU at 100% with FR running	Server-side FR mode on too many channels without GPU	Switch to Edge FR mode on WizMind cameras, or install a compatible GPU
FR works for clean faces but fails with masks or glasses	Occlusion filter set too strictly, or enrollment photos don’t include occluded versions	Enroll additional photos showing the person with glasses if applicable; adjust occlusion filter tolerance
Bulk import fails for some records	Photo file name mismatch or unsupported image format	Ensure photos are JPEG (not PNG or HEIC), and file names exactly match the ID column in the template

---

Performance and Accuracy Best Practices

• Use edge FR whenever possible. WizMind cameras offload processing to the device, keeping server load flat even as you scale to 20+ FR channels.
• Enroll multiple photos per person. 3–5 photos from slightly different angles and lighting conditions can increase recognition rate by 15–20% in real-world conditions compared to a single enrollment photo.
• Keep the person library clean. Outdated records, poor-quality enrollment photos, and duplicate persons all degrade matching performance. Audit the library quarterly on active deployments.
• Set thresholds per group, not globally. A blocklist of banned persons warrants a lower threshold (more sensitive) than a general staff allowlist. Use DSS Pro’s per-group threshold to tune this separately.
• Use the stranger alert sparingly. Stranger alerts (unrecognized face) are extremely high volume in any public-facing environment. Reserve this for controlled-access areas where every entrant should be in the database.
• Schedule FR processing during off-peak hours for historical search. Running deep face searches over weeks of archive footage is compute-intensive. Schedule batch searches during overnight hours to avoid impacting live system performance.
• Monitor the FR event queue. Navigate to Maintenance → System Status → AI Queue to verify the FR processing queue is not backing up. A persistent queue backlog means the server is under-provisioned for the current FR channel count.

---

DSS Pro FR Capacity Reference

Parameter	DSS Professional 8.7
Max persons in library	100,000
Max face photos per person	10
FR channels in base license (DSS8PRVB)	2
Max FR channels (with add-on licenses)	Limited only by server capacity
Minimum face size for detection	80 × 80 px (configurable)
Similarity threshold range	50% – 100% (default: 85%)
Face snapshot retention	Configurable (30–365 days or unlimited)
Person groups per system	Unlimited
Access control integration	✅ Via Dahua FR terminals (face sync)

---

💬 Need a license or have questions? → Message us on Telegram — free consultation, usually reply within a few hours.

Frequently Asked Questions

How many FR channel licenses do I need for a deployment with 50 cameras?

You only need an FR license for cameras that will actively perform face recognition. If only 8 of your 50 cameras are at choke points where FR is useful, you need 8 FR channel licenses (6 add-ons beyond the 2 in the base license). The remaining 42 cameras operate normally without FR without requiring any additional licensing.

Can DSS Pro do facial recognition on recorded footage, or only live video?

Both. DSS Pro 8.7 stores face capture events from live video in its FR database, enabling face-based search over any recorded period. You can upload a reference photo and search across all FR cameras for any time range. This forensic retrieval capability is one of the most practically useful features for post-incident investigation.

What is the difference between Face Detection and Face Recognition in DSS Pro?

Face Detection simply identifies that a human face is present in the frame and captures a snapshot — no matching against a database. Face Recognition takes the captured face and compares it against the Person Library to find a match. Detection is included in basic camera analytics; Recognition requires the FR license and Person Library configuration described in this guide.

Does DSS Pro facial recognition work with non-Dahua cameras?

Server-side FR mode works with any camera that can deliver a clear face image (720p minimum, ONVIF or RTSP stream). Edge FR mode requires Dahua WizMind AI cameras. The practical limitation is that AI metadata (face bounding boxes, quality scores) from non-Dahua cameras is not available, so server-side processing handles everything — which is functional but more resource-intensive and slightly less accurate than edge FR on WizMind hardware.

How do I handle persons who are no longer authorized (e.g., terminated employees)?

Navigate to Person Management, find the person record, and either delete it or move it to a different group (e.g., from Staff Allowlist to a Terminated/Blocklist group). Changes propagate immediately to all FR channels linked to the affected groups. If the person is also enrolled in a Dahua FR access terminal, the sync will push the deletion to the terminal automatically.

What similarity threshold should I use as a starting point?

Start at 85% for most deployments. Run for 2–3 days, review the false positive and false negative rates, and adjust. Move to 82–83% if you are missing too many genuine matches; move to 87–88% if you are seeing too many incorrect alerts. Avoid going below 78% — at that level, similarity scores become unreliable and false positives will overwhelm operators.

Can multiple persons be matched simultaneously in a single camera frame?

Yes. DSS Pro performs FR on all detected faces in the frame simultaneously. In a crowded lobby scene, the system can match several persons at the same time, generating individual match records for each. Performance scales with server capacity; on edge FR cameras, simultaneous multi-face matching is handled on-device.

---

Summary

Setting up facial recognition in DSS Pro 8.7 correctly requires attention to each stage of the pipeline: right license, right camera protocol, correct detection zone, well-populated person library, per-group similarity thresholds, and properly targeted alert rules. Skip any one of these steps and the system will be technically running but practically unreliable.

The single highest-impact action for any FR deployment is camera placement — a WizMind camera at the right angle and height, with adequate front lighting, will outperform an expensive server-side FR setup on a badly placed camera every time. Get the hardware right first, then configure the software around it.

 

---