A dedicated solution for following athlete movements in real time—even in complex, high-speed action scenes.

This model detects players and maintains consistent tracking IDs over time, returning structured track outputs you can use for advanced performance metrics, coaching tools, and enriched sports broadcasting experiences.

Use it on images, recorded video, or live streams. No custom training required.

Optimized for:

• Multi-player tracking with persistent IDs
• Fast motion, partial occlusion, and camera movement
• Frame-by-frame track outputs for video
• Cloud or On-Prem deployment
• Rapid setup for prototype → production
  ‍

Why This Model Exists

Detection tells you what’s in a frame.

Tracking tells you what happened over time.

Sports analytics breaks the moment you rely on detection alone:

• Players cross paths and swap positions
• Occlusions happen constantly (refs, teammates, motion blur)
• Broadcast cameras pan/zoom and change angle
• “Count players” is easy — “follow player #12 for 30 seconds” is not

Most teams end up building their own tracking stack on top of a detector. That work is deceptively expensive:

• ID switching makes metrics unreliable
• Small tracking errors compound into bad stats
• Debugging is painful in chaotic action sequences
• Different sports and camera styles require lots of tuning

This model exists to provide a ready-to-use baseline:
player tracks with consistent IDs, so you can jump straight into metrics, coaching insights, and broadcast features—without assembling (and maintaining) a full tracking pipeline from scratch.

‍

Key Capabilities

Input Types

• Video files
• RTSP / livestream feeds
• Webcam / IP camera streams
  (Tracking is primarily intended for video/streams.)

Output

• JSON with per-frame detections + persistent track IDs
• Bounding boxes per player
• Confidence scores
• Track metadata (start/end frames, durations where supported)

Setup

• Create account
• Get API key
• Send video or stream
• Receive structured tracks in real time

No training. No labeling. Minimal configuration.

‍

Example Output

{
  "tracks": [
    {
      "track_id": 12,
      "label": "player",
      "confidence": 0.93,
      "frames": [
        {
          "frame_index": 184,
          "timestamp_ms": 6133,
          "bbox": { "x": 214.6, "y": 188.4, "width": 312.1, "height": 708.9 },
          "confidence": 0.95
        },
        {
          "frame_index": 185,
          "timestamp_ms": 6166,
          "bbox": { "x": 221.2, "y": 186.9, "width": 311.7, "height": 709.3 },
          "confidence": 0.94
        }
      ]
    },
    {
      "track_id": 27,
      "label": "player",
      "confidence": 0.91,
      "frames": [
        {
          "frame_index": 184,
          "timestamp_ms": 6133,
          "bbox": { "x": 812.3, "y": 205.6, "width": 298.4, "height": 692.5 },
          "confidence": 0.93
        }
      ]
    }
  ],
  "source_width": 1920,
  "source_height": 1080
}

(Adjust fields to match your exact tracking schema: whether tracks are grouped by ID, returned per-frame, include velocity, etc.)
‍

Practical Use Cases

Performance Metrics & Sports Analytics

• Distance covered, speed, acceleration (with calibration)
• Time in zones / formations
• Player movement heatmaps
• Possession / involvement cues (paired with ball tracking)

Coaching Tools

• Drill and movement consistency analysis
• Player positioning review over time
• Automated clip creation per player ID
• Tactical pattern visualization

Broadcasting & Production Enhancements

• Live overlays that follow a player
• Auto-generated “player spotlight” replays
• Metadata for quick search (“show clips where player 12 sprinted”)
• Enhanced highlight workflows and tagging

Fan Engagement

• Second-screen player tracking views
• Personalized highlights (“follow my favorite player”)
• Interactive overlays and stats triggers

‍

Why This Output Matters

Tracking unlocks time-based understanding:

• Identity: the same player stays the same track ID
• Continuity: movement is measurable, not just detected
• Metrics: you can compute speed, distance, and patterns
• Automation: you can trigger highlights and overlays reliably

If your product depends on athlete movement—not just detection—tracking is the core primitive.

‍

Deployment Options

EyePop Cloud

• Scalable
• Managed infrastructure
• Best for web apps + fast iteration
  

On-Premise Runtime

• Keep video inside your network
• Lower latency options
• Works with GPU or CPU environments
• Ideal for regulated or sensitive environments
  
  

Who This Is For

• Sports analytics platforms and performance teams
• Coaching tools and training systems
• Broadcasters and production teams building overlays
• Fan engagement products that need time-based player context