Detector24
Back to Model Catalogue
Wanted Persons Detection
VideoFace / People Related

Wanted Persons Detection

Real-time face recognition against 52+ law enforcement databases including Interpol, Europol, and FBI. Video surveillance watchlist screening.

Accuracy
97.5%
Avg. Speed
5.0s
Per Minute
$0.2500
API Name
wanted-person-detection

Bynn Wanted Persons Detection (Video)

The Bynn Wanted Persons Detection model performs real-time face recognition in video streams against comprehensive law enforcement databases including Interpol Red Notices, Europol Most Wanted, FBI Most Wanted, and 52 country-specific watchlists. This model enables continuous monitoring of video feeds to identify fugitives, missing persons, and individuals subject to international arrest warrants.

The Challenge

Wanted criminals cross borders. Fugitives create new identities. Missing persons appear in unexpected locations. Traditional law enforcement relies on officer memory and manual database queries—a wanted person walks through an airport checkpoint while their warrant sits unflagged in a system no one thought to check.

The databases are massive and constantly updating. Interpol's Red Notices, national most-wanted lists, missing persons reports, and abduction alerts change hourly. A person reported missing in one country may appear on CCTV footage in another. A fugitive from justice may attempt border crossing using fraudulent documents. The window for apprehension is brief—by the time manual verification completes, the person has moved on.

KYC and onboarding workflows face similar challenges. Financial institutions, cryptocurrency exchanges, and regulated businesses must screen customers against watchlists. A single missed match can result in regulatory penalties, facilitating money laundering, or enabling sanctioned individuals to access financial systems. But screening millions of customer photos against constantly evolving watchlists manually is impossible.

The stakes are highest for abducted persons and missing children. Every hour counts. CCTV cameras capture faces continuously, but human monitors cannot compare every face against missing persons databases. Automated recognition can flag potential matches instantly, enabling rescue operations that would otherwise arrive too late.

Video surveillance generates far more data than static images—hours of footage from multiple cameras require real-time analysis. A wanted person may appear for seconds in a crowd. Missing that moment means missing the opportunity for intervention.

Model Overview

The Bynn Wanted Persons Detection model performs continuous facial recognition in video streams, analyzing frames to detect and match faces against a database refreshed every 3 hours with new wanted persons, missing individuals, and abduction alerts from international law enforcement agencies.

Achieving 97.5% accuracy, the model processes video in real-time, identifying potential matches as they appear and providing timestamps for efficient review and response.

How It Works

The model performs comprehensive video analysis with facial recognition:

  • Frame extraction: Processes video frames at optimal intervals for face detection
  • Face tracking: Identifies and tracks faces across video frames
  • Feature extraction: Generates facial embeddings for comparison
  • Real-time matching: Compares detected faces against watchlist database updated every 3 hours
  • Temporal localization: Provides precise timestamps when matches appear
  • Multi-database search: Searches across Interpol, Europol, FBI, and 52 country databases simultaneously

Response Structure

The API returns a structured response containing:

  • matches: Array of potential watchlist matches, each containing:
    • person_id: Unique identifier in the watchlist database
    • name: Name associated with the watchlist entry
    • confidence: Match confidence score (0.0-1.0)
    • database: Source database (Interpol, FBI, Europol, etc.)
    • case_type: Wanted, missing, abducted, or other classification
    • timestamps: Array of video timestamps where the person appears (mm:ss.ff format)
    • frame_count: Number of frames where the match was detected
  • num_faces_detected: Total number of distinct faces found in the video
  • num_matches: Number of watchlist matches identified

Covered Databases

Database Coverage Update Frequency
Interpol Red Notices International arrest warrants from 195 member countries Every 3 hours
Europol Most Wanted Europe's most wanted fugitives Every 3 hours
FBI Most Wanted United States federal fugitives and missing persons Every 3 hours
National Watchlists 52 country-specific law enforcement databases Every 3 hours

Performance Metrics

Metric Value
Recognition Accuracy 97.5%
Average Response Time 5,000ms
Database Update Frequency Every 3 hours
Max File Size 100MB
Supported Formats MP4, MOV, AVI, WebM, MKV

Use Cases

  • Airport Security: Monitor security camera feeds at checkpoints and terminals for wanted persons and missing children
  • Border Protection: Screen travelers at border crossings in real-time
  • KYC Video Verification: Screen customer video submissions against watchlists during onboarding
  • Public Surveillance: Monitor CCTV networks in transportation hubs, public spaces, and critical infrastructure
  • Missing Persons Recovery: Identify abducted children and missing persons from continuous video surveillance
  • Event Security: Monitor live video feeds at concerts, sporting events, and public gatherings
  • Transit Systems: Screen passengers on trains, buses, and ferries
  • Casino Surveillance: Identify banned or wanted individuals entering gaming facilities

Known Limitations

Critical Considerations:

  • Not Definitive Identification: Matches indicate similarity, not guaranteed identity; human verification required
  • Age & Appearance Changes: Watchlist photos may be years old; subjects may have changed appearance significantly
  • Video Quality: Poor lighting, low resolution, motion blur, or partial face visibility reduce match accuracy
  • False Positives Possible: Innocent individuals may resemble wanted persons; verification protocols essential
  • Brief Appearances: Very short appearances or rapid movement may reduce detection confidence
  • Database Coverage: Limited to included law enforcement databases; regional databases may not be comprehensive

Legal & Ethical Considerations

⚠️ This tool must be used responsibly and in compliance with applicable laws.

Critical Requirements

  • Verification Protocols: All matches require human verification before law enforcement contact or action
  • CCTV Usage Authorization: Check with local authorities regarding permissions and requirements for using facial recognition on CCTV systems in your jurisdiction
  • KYC Regulatory Requirement: For financial services and regulated industries, watchlist screening is typically a legal requirement, not optional
  • Due Process: Matches do not constitute probable cause; follow legal protocols for person-of-interest handling
  • Data Retention: Implement appropriate retention policies for facial recognition data in compliance with local regulations
  • Recording Consent: Comply with video surveillance and recording consent laws in your jurisdiction

Operational Protocols

  • Trained Personnel: Only trained security or compliance personnel should review matches
  • Escalation Procedures: Establish clear protocols for handling confirmed matches
  • Law Enforcement Coordination: Coordinate with appropriate authorities; do not attempt apprehension
  • False Positive Response: Have protocols to avoid wrongful detention or harassment
  • Emergency Response: For missing/abducted persons, establish rapid response procedures with law enforcement

Disclaimers

  • Not Law Enforcement: This is a screening tool for authorized use; does not grant law enforcement authority
  • Probability-Based Match: A flagged match indicates high probability, NOT confirmed identity. Local authorities must conduct additional verification checks before taking any action
  • No Match Does Not Clear: Absence of a match does not mean a person is not wanted. Database entries may have been removed, photos may be outdated, or detection may have been inconclusive due to video quality
  • Database Limitations: Not all wanted persons are in covered databases; regional or local warrants may not be included
  • Professional Use Only: Requires proper training, legal authority, and operational protocols
  • Regulatory Compliance: Users are responsible for compliance with all applicable laws and regulations

Best Practice: Use video wanted persons detection for continuous monitoring with real-time alerts to trained security personnel. Implement strict verification procedures and maintain coordination with law enforcement agencies for confirmed matches.

API Reference

Version
2601
Jan 3, 2026
Avg. Processing
5.0s
Per Minute
$0.25
Required Plan
trial

Input Parameters

Detect wanted persons (Interpol Red/Yellow Notices) in video or batch images. Provide either video (video_url/base64_video) OR images (image_urls/base64_images), not both.

video_urlstring

URL of video to scan for wanted persons

Example:
https://example.com/video.mp4
base64_videostring

Base64-encoded video data (alternative to video_url)

image_urlsarray

Array of image URLs for batch processing

Example:
["https://example.com/1.jpg","https://example.com/2.jpg"]
base64_imagesarray

Array of base64-encoded images (alternative to image_urls)

fpsinteger

Frames per second to extract from video (1-10)

Example:
4
distance_thresholdfloat

Face matching distance threshold (0.3-0.7, lower = stricter)

Example:
0.5
extract_facesboolean

Extract cropped face PNG for best match of each person

Example:
true

Response Fields

Wanted person detection results with matches and processing statistics

matchesarray

Array of matched wanted persons

Array Item Properties:
matched_personobject

Information about the matched wanted person

best_matchobject

Details of the highest confidence detection

all_detectionsarray

All frames/images where this person was detected

occurrence_countinteger

Number of frames/images where person was detected

statsobject

Processing statistics

Object Properties:
input_typestring

Type of input (video or images)

items_processedinteger

Number of frames/images processed

faces_detectedinteger

Total faces detected across all frames/images

unique_matchesinteger

Number of unique wanted persons matched

processing_time_msinteger

Total processing time in milliseconds

Complete Example

Request

{
  "model": "wanted-person-detection",
  "video_url": "https://example.com/surveillance.mp4",
  "fps": 4,
  "distance_threshold": 0.5,
  "extract_faces": true
}

Response

{
  "inference_id": "inf_wpd_abc123def456",
  "model_id": "wanted_person_detection",
  "model_name": "Wanted Person Detection",
  "moderation_type": "video",
  "status": "completed",
  "result": {
    "matches": [
      {
        "matched_person": {
          "token": "face_xyz789",
          "name": "John Doe",
          "source": "wanted_persons",
          "metadata": {
            "nationalities": [
              "US"
            ],
            "date_of_birth": "1985-03-15",
            "gender": "male"
          }
        },
        "best_match": {
          "index": 24,
          "timestamp_ms": 6000,
          "bbox": {
            "x1": 150,
            "y1": 80,
            "x2": 280,
            "y2": 240
          },
          "det_score": 0.94,
          "distance": 0.18,
          "similarity": 0.82,
          "face_crop_png": "<base64-encoded-png>",
          "crop_width": 156,
          "crop_height": 192
        },
        "all_detections": [
          {
            "index": 20,
            "timestamp_ms": 5000,
            "bbox": {
              "x1": 145,
              "y1": 82,
              "x2": 275,
              "y2": 238
            },
            "det_score": 0.91,
            "distance": 0.22,
            "similarity": 0.78
          },
          {
            "index": 24,
            "timestamp_ms": 6000,
            "bbox": {
              "x1": 150,
              "y1": 80,
              "x2": 280,
              "y2": 240
            },
            "det_score": 0.94,
            "distance": 0.18,
            "similarity": 0.82
          },
          {
            "index": 28,
            "timestamp_ms": 7000,
            "bbox": {
              "x1": 160,
              "y1": 85,
              "x2": 290,
              "y2": 245
            },
            "det_score": 0.89,
            "distance": 0.25,
            "similarity": 0.75
          }
        ],
        "occurrence_count": 3
      }
    ],
    "stats": {
      "input_type": "video",
      "items_processed": 120,
      "faces_detected": 45,
      "unique_matches": 1,
      "processing_time_ms": 8500
    }
  },
  "response_time_ms": 8500,
  "created_at": "2026-02-07T10:47:57Z",
  "completed_at": "2026-02-07T10:48:05Z"
}

Additional Information

Rate Limiting
If we throttle your request, you will receive a 429 HTTP error code along with an error message. You should then retry with an exponential back-off strategy, meaning that you should retry after 4 seconds, then 8 seconds, then 16 seconds, etc.
Supported Formats
mp4, mov, avi, webm, mkv
Maximum File Size
100MB
Tags:face-recognitionwanted-personsinterpolsecuritylaw-enforcementwatchlist

Ready to get started?

Integrate Wanted Persons Detection into your application today with our easy-to-use API.

Learn more about video moderationBrowse all models