Deeply’s Listen AI Industrial uses acoustic analysis to catch connector engagement defects on auto assembly lines, distinguishing first- and second-lock clicks even amid air-gun and metal-friction noise. The foundation model, trained on 2.1 million hours of factory audio, is effective from day one and has achieved 99.87% accuracy at a global automaker’s plants in Korea and Mexico, driving interest from North American OEMs. Deeply will showcase live demos and consultations at Automate 2026 in Chicago, June 22–25, as it expands deployment across global production lines.
How do you see acoustic AI complementing or replacing vision-based quality checks on noisy, fast-moving assembly lines?
-Deeply's 'Listen AI Industrial' has already proven to have 99.87% accuracy in global automaker (Company H) production lines in Korea and Mexico.
-Foundation model trained on 2.1 million hours of factory noise perfectly identifies subtle 'engagement sounds' from the first day of deployment.
-Distinguishes 1st and 2nd locking sounds even amidst air gun and metal friction noises, driving North American OEM adoption
-Live demo and consultation of connector engagement diagnostic solution at Automate 2026 in Chicago, June 22-25.
We’ve been testing acoustics for snap-fit cap verification at 300+ ppm where air knives and pneumatics keep shifting the noise floor; how sensitive is your model to drift from air leaks or tool wear, and do you report false negatives vs false positives separately (99.87% can mask imbalance)? For controls, can your edge publish a deterministic pass/fail bit over EtherNet/IP to a PLC in under 100 ms, and what tuning is needed when connector types change?