As demand grows for AI solutions that process images from cameras and other photo sensors, more efficient systems are needed. Like NVIDIA and others before it and drawing on its mobile technology developments, chip maker Qualcomm has unveiled an R&D project that reduces the compute power needed to handle visual AI, in turn reducing the size, price and power consumption of the processors used.
The R&D team’s methodology eliminates redundancy by only analysing frame differences rather than each frame and skipping frames where no changes are evident. This creates faster processing that understands the relationship between frames. Less complex, smaller, less expensive processors mean smaller, less power-hungry finished devices, such as smarter cameras with embedded AI tech, for example.
Also, for hybrid, distributed infrastructures, in the cloud or at the network edge, this leads to less costly, faster data analysis with less latency and reduced data sharing, which improves security
Along with other big names like Intel/Moviidius, Google, Microsoft, Samsung, NXP and ARM, big steps are being made to produce much lower-cost visual AI technologies to make everybody’s lives, including those of content creators, that little bit easier while reducing costs.