Product Updates
-
System Update 0.1.4
Improved alert filtering to reduce unnecessary system noise Early versions of the system prioritised detection sensitivity. This update refines alert thresholds to better distinguish between low-risk observations and events that require attention. • Refined alert triggering thresholds • Improved detection confidence scoring • Reduced unnecessary safety alerts
-
System Update 0.1.3
Improved worker detection stability in high-activity scenes Active construction areas can involve multiple workers moving simultaneously. This update improves detection stability and reduces duplicate detections during complex movement patterns. • Improved frame-to-frame tracking consistency • Reduced duplicate detection events • Enhanced worker boundary detection
-
System Update 0.1.2
Improved PPE detection performance in dusty environments Dust and particulate matter are common challenges on construction sites. This update improves model robustness under reduced visibility conditions, helping maintain detection reliability during active work phases. • Expanded dataset with real construction site footage • Improved low-visibility detection robustness • Reduced reflective surface false positives
-
System Update 0.1.1
Improved helmet detection under partially obstructed conditions Construction environments often involve overlapping workers and temporary visual obstruction. This update improves the model’s ability to correctly detect helmets when workers are partially blocked by equipment or other personnel. • Improved worker segmentation performance • Improved worker segmentation performance • Reduced missed detections in multi-person scenes
-
System Update 0.1.0
Initial PPE detection capability deployed for construction environments This update introduces the first operational version of the PPE detection system designed for dynamic construction sites. The system establishes the baseline capability for identifying essential protective equipment in real time. • Baseline AI model trained for helmet and vest recognition • Real-time object detection pipeline implemented
