Summary：Hanwha Vision and GIST claim DCASE victory with unforgettable audio AI for security cameras **Intro

Hanwha Vision and GIST claim DCASE victory with unforgettable audio AI for security cameras

**Introduction**
Hanwha Vision, a global leader in video surveillance, teamed up with the Gwangju Institute of Science and Technology (GIST) to clinch the top prize at the recent Detection and Classification of Acoustic Scenes and Events (DCASE) challenge. Their joint entry showcased an audio‑AI system that can pinpoint abnormal sounds—such as glass breaking, shouting, or gunfire—in real time, feeding alerts directly to security‑camera feeds. The win underscores how sound‑based intelligence is becoming a critical layer in modern surveillance ecosystems.

**Key Developments**
The winning model combines a lightweight convolutional neural network with a transformer‑based temporal encoder, optimized for edge deployment on Hanwha’s latest AI‑enabled cameras. By training on a diverse dataset that mixes urban, indoor, and outdoor acoustic scenes, the system achieves a mean average precision of 0.92 on the DCASE test set while consuming less than 150 mW of power. Hanwha integrated the algorithm into its Wisenet AI platform, allowing existing hardware to receive firmware upgrades without additional sensors. GIST contributed novel data‑augmentation techniques that improve robustness against reverberation and background noise, a persistent challenge in real‑world installations.

**Industry Analysis**
Security operators have long relied on visual analytics, yet audio cues often precede visible incidents. The DCASE victory signals a shift toward multimodal perception, where sound and video are processed jointly to reduce false alarms and accelerate response times. Market