Home > other > YOLO in ARM speed optimization scheme
YOLO in ARM speed optimization scheme
Time:11-30
Recently the YOLO algorithm to use the ARM on the run, does not require real-time, but at least can hope to detect time within 1 s, I will be the original YOLO on ARM to run more than 42 s, beg god give directions! If the yolo on caffe on a move to the ARM will be faster?
CodePudding user response:
Their top, ha, ha, ha
CodePudding user response:
YOLO not several versions of the model is the smallest Tiny YOLO also want 2 s? https://pjreddie.com/darknet/yolo/
CodePudding user response:
If you have any GPU is running under the PC 2 s, but I want to run into the ARM, it is good to slow
CodePudding user response:
Ported to arm have what resources? I want to try to transplant, too
CodePudding user response:
ARM is generally consider Neon instruction optimization? But then again, now tencent and baidu isn't already open source framework on the arm, his transplant PC code, a lot of work
CodePudding user response:
Neural network model to the mobile phone is now in the job, I suggest you try the model pruning, quantification of these two methods, of course, can have certain influence on the accuracy, the need to balance the two aspects,