Microsoft Research330 тыс
Опубликовано 17 августа 2016, 2:23
In order to fulfill todayΓÇÖs object recognition requirements in mobile environment, the brain-inspired Unified Visual Attention Model (UVAM) is applied to SIFT-based recognition for high processing speed with limited energy budget. In addition, a heterogeneous multicore processor is proposed using SIMD/MIMD/ASIC combined architecture for different data-level parallelism of proposed algorithm. Especially, we integrated an artificial intelligence accelerator for inference and learning operation to enable smart object recognition system for future video applications.