Extract individual person locations from crowd density maps

Use This When

• Converting density-based crowd counts into discrete person coordinates
• Building tracking systems that need initialization points for crowded scenes
• Creating visualizations that mark individual person positions from density estimates
• Implementing zone-based analytics that count people per region using peak locations

What It Does

• Finds local maxima in density map using maximum filter with configurable window size
• Identifies top N peaks based on density values where N equals total person count
• Returns center-of-mass coordinates for each detected peak as point list
• Filters peaks to prevent double-counting individuals in overlapping regions

Works Best With

• estimate-crowd-count → this component → visualize for person position overlays
• Peak locations → zone containment analysis for per-region crowd distribution
• Integration with tracking systems to seed trackers in dense initialization scenarios
• Feeding analytics dashboards that need individual counts per spatial region

Caveats

• Window size critically affects peak separation; too small causes duplicate detections
• Assumes each peak represents one person; fails when density map has merged responses
• Peak ordering by density value may not reflect spatial importance in all scenarios
• Count mismatch between input and detected peaks indicates density map quality issues