In this episode, Lucas and Luna dive into the on-device AI behind real-time background blur on smartphones. They explore how Apple, Google, and Samsung implement portrait mode for video calls and live streaming, using neural networks and depth sensors. The discussion covers the evolution from dual-camera setups to single-camera solutions with machine learning, the role of the Neural Engine and Tensor Processing Unit, and how apps like Zoom and Instagram leverage these features. Specific numbers include the 15 billion operations per second required for real-time blur at 30 frames per second on a flagship chip. Lucas explains how the model is trained on synthetic data and fine-tuned for edge devices, while Luna questions the trade-offs in accuracy and power consumption. The episode also touches on the future of computational photography and the impact on privacy, since processing happens entirely on-device. No prior episode has covered this specific angle of on-device AI for real-time background blur in mobile apps.

#OnDeviceAI #BackgroundBlur #PortraitMode #ComputationalPhotography #MobileApps #Apple #Google #Samsung #NeuralEngine #TensorProcessingUnit #Zoom #Instagram #DepthMapping #MachineLearning #RealTimeVideo #SmartphoneCamera #Technology #FexingoBusiness

Keep every episode free: buymeacoffee.com/fexingo