Acceleration And Energy Reduction Of Object Detection On Mobile Graphics Processing Unit

The evolution of high performance computing in today’s smartphones is enablingrntheir use in compute-intensive applications. As the compute requirement increases,rnthe energy required to do the computation cannot increase in proportion becausernthe cost of providing that energy available and cooling would become prohibitive.rnAn alternative, potentially power-reducing approach is to use graphics processingrnunits or special accelerator cores. Today’s smartphones are equipped with systemon-chiprn(SoC) devices that house many cores such as graphics processing units, digitalrnsignal processors, and special multimedia encoder/decoder hardware along sidernmulti-core central processing units. Their inclusion enables applications that requirerngreater computational power such as real-time computer vision. In this work, wernstudy the capability of the recently introduced general-purpose graphics processingrnunit (GPU) in a smartphone SoC to enable energy-efficient object detection. Thisrnwill include understanding the architecture of the recent GPUs that will be used (thernAdreno 320 and Adreno 420 from Qualcomm), the implementation and optimizationrnof the object detection algorithm used in the Open Computer Vision libraryrn(OpenCV) using these GPUs and measuring the energy consumption of this implementation.rnWe implemented the Viola-Jones based object detection on the GPU inrnan Android tablet. The implementation is 35% faster on average than the same algorithmrnrunning on the CPU on the same device. The implementation also reduces thernaverage energy consumption by 68% compared to the CPU on the same device. Anrnapplication that utilized the object detector on the mobile GPU to detect Ringwormrnskin disease was developed. A classifier was trained for this application and it hasrnan accuracy of 75%.

Get Full Work

Report copyright infringement or plagiarism