User Guide

Smartag is an AI-powered platform for analyzing sports videos. With YOLO object detection and customizable label mapping, you can quickly identify athletes, equipment, and game moments.

1. Visit the Analyzer Dashboard
2. Upload your sports video or image
3. Configure label mappings for your analysis
4. Click "Analyze with AI" to run detection
5. Download results as PNG or JSON

Smartag supports both images and videos. You can upload files in several ways:

• Drag & Drop: Drag media files directly onto the upload zone
• Click Upload: Click the upload area to browse your computer
• Batch Processing: Upload ODS files with image URLs for batch analysis

Label mapping is a powerful feature that lets you convert YOLO detection labels (like "person") into your own custom sports analytics tags (like "athlete", "player", "position").

How to Create a Mapping:

1. Enter a System Label (from YOLO detection, e.g., "person")
2. Enter Custom Tags separated by commas (e.g., "athlete, player, visible")
3. Click Add Mapping
4. Your mapping is saved in browser storage and persists across sessions

Example Mappings:

Pro tip: Use many-to-many mapping to create flexible analytics. One YOLO label can map to multiple custom tags, giving you detailed categorization of detected objects.

Once you've uploaded media and configured your label mappings, you're ready to analyze.

Analysis Steps:

1. Upload your image or video file
2. Review and adjust label mappings as needed
3. Click "Analyze with AI" button
4. Wait for YOLO detection to complete
5. Review detected objects in the results table
6. Download annotated image or JSON report

Understanding Results:

• Label: The YOLO detection label
• Tags: Your custom mapped tags for this label
• Confidence: How confident YOLO is in this detection (0-100%)
• Bounding Box: Pixel coordinates [x1, y1, x2, y2]

After analysis, use the tag filter to narrow down results to specific objects of interest.

• Type a tag name in the Filter by tag field
• Results automatically update to show only matching detections
• Useful for focusing on specific athletes, equipment, or game elements

Smartag provides two formats for results:

Annotated PNG

A visual representation of your image/video frame with bounding boxes and labels drawn on top. Perfect for visual review and presentations.

JSON Report

A structured data format containing all detection details:

{
  "label": "person",
  "tags": ["athlete", "player"],
  "confidence": 0.94,
  "bbox": [100, 50, 400, 500]
}

Use JSON reports for integration with external analytics tools or data processing pipelines.

• Larger images take longer to analyze - consider resizing if needed
• Save your label mappings by using the mapping panel - they persist across sessions
• Use specific tags in your mappings for better filtering capabilities
• Confidence threshold typically ranges from 0.85-0.95 for good results
• For batch processing, consider splitting large datasets into multiple runs