Both challenges are taking place now until early March 2021, and aim to obtain a solution with the best trade-off between prediction accuracy and the latency on MediaTek Dimensity 1000+ platform.
Edge AI hardware and software deployment is increasingly in demand, through the requirement of high performance, power-efficient AI functionality in a wide range of applications. MediaTek develops its own in-house deep learning accelerator technologies that feature across its product portfolio including mobile, home and IoT/Enterprise.
CVPR’21 – Mobile AI Challenge: Mobile AI 2021 Learned Smartphone ISP Challenge
– Challenge information
The Mobile AI workshop is the first ever CVPR challenge to evaluate deep learning models on an application specific, mobile neural network accelerator instead of general purpose CPU and GPU hardware. ETH Zurich (ETHZ) and MediaTek have together organized the challenge, aimed at novel Computer Vision opportunities , with access to MediaTek Dimensity 1000+ hardware.
Participants can submit their model online to test on the Dimensity 1000+ (MediaTek APU 3.0, 4.5 TOPS heterogeneous multi-processor) device farm at any time during the competition period.
A total prize pool of USD$3000 is available for those entering the challenge.