At NTT Upgrade 2025 in San Francisco, NTT Research has announced a UHD Video AI Inference chip for real-time 4k video processing. Its first application for the technology is for use in drones doing surveillance and detecting people flow. The company claims it will detect people using a 3×21 pixel pattern from a drone at 500 feet. Kazuhiro Gomi, President and CEO of NTT Research, said, “The AI Inference chip can be embedded into a camera and applied to high-resolution video. When you apply it to the video, you can detect objects in the image. “Conventional technology is limited due to the computational need. You have to downgrade the quality of the video or lower the resolution of the camera. Its power requirements limit its use on devices.” The chip breaks each image down into multiple pieces. The AI inference engine then works on each piece in parallel. Parallel processing allows it to detect individual objects or even parts of an object. Once it has finished, it reassembles the image, having identified all of the objects. What is especially important is that this is working in real-time on 4k video at 30 frames per second.

Full report : NTT Throws Hat Into AI Chip Race and Targets Drone Footage Tech.