AI models trained on in-car camera footage preserve important details such as vehicles, lanes, and pedestrians while compressing less important elements to reduce file size without compromising evidence quality.LightMetrics, an AI-powered video telematics company, on Thursday announced the introduction of its proprietary Neural Network Video Compression (NNVC) technology in India, an AI-powered software solution that enables commercial vehicles to store up to 2.5x more on-board camera footage on existing hardware while reducing video upload costs by up to 50%.
According to an official press release, the software is designed to address rising storage costs due to rising global prices for DRAM and flash memory. NNVC uses AI to maximize the storage efficiency of existing vehicle camera systems, instead of requiring larger SD cards or hardware upgrades.
The technology uses deep neural networks trained specifically on in-vehicle camera footage to preserve important visual information such as vehicles, lane markings, pedestrians, and driver behavior while compressing less important scene elements to reduce file size without compromising evidence quality.
Citing a Frost & Sullivan white paper, LightMetrics said AI-based compression delivers approximately 3-5x better compression ratios than traditional H.264 compression, while extending video retention on devices, reducing connectivity costs, and enabling faster retrieval of incident footage.
“NNVC is designed to solve this challenge with AI, allowing fleets to leverage their existing hardware to significantly increase storage capacity and reduce data costs,” said Soumik Ukil, co-founder and CEO of LightMetrics.
The technology is already being deployed across customer fleets, and initial implementations have demonstrated increased storage efficiency and reduced data consumption, the company said.
LightMetrics expects AI-powered video compression to help fleet operators scale video telematics more cost-effectively while enhancing safety, compliance, and operational efficiency.
