Deefly Listen AI
3 steps1 pieces5 Sources
South Korean startup Deefly (Deeply Inc.) has reported that its acoustic AI system, 'Listen AI', achieved a 99.9% anomaly detection rate with zero false alarms during a deployment at a PSA International-operated port terminal in South Korea. The announcement was made by CEO Lee Suji at the 'PSA Group Innovation Festival 2026' in Singapore on November 16, 2026.
The system monitors transfer cranes used to move containers. By learning normal operational sounds and identifying deviations, it enables predictive maintenance without requiring physical sensor attachment to existing equipment. In a one-week test collecting over 4,700 audio samples, the AI correctly identified anomalies. A specific case involved a motor where the anomaly score doubled compared to normal levels; after pre-emptive repair, the score dropped by 34%, confirming the fix.
This development matters because industrial acoustic AI faces challenges in noisy environments and data scarcity. Traditional methods often rely on manual listening or simple thresholds, leading to inconsistent results. Deefly's approach uses deep learning to filter ambient noise, claiming high accuracy even in 100-decibel environments. However, the industry remains divided on whether such non-intrusive, sound-based monitoring can universally replace vibration sensors or visual inspections for all critical infrastructure. While Deefly cites success in ports, semiconductors, and rail, competitors like Amazon Monitron emphasize multi-modal sensing (vibration/temperature) for robustness.
Readers should watch for: 1) Deefly's expansion of Listen AI to other PSA international terminals, currently under discussion; 2) Verification of the 'zero false alarm' claim in longer-term deployments beyond the initial one-week window; 3) Adoption rates among global manufacturers who have previously relied on traditional condition monitoring systems.
How it got here
2024-11-13: Deefly introduces 'Listen AI' as an AI-powered sound analysis safety solution, highlighting capabilities in detecting over 100 distinct sounds and integration with CCTV/dashcams.
deeplyinc.com ↗