Parallel acceleration for detection models
Closed this issue · 6 comments
@LewisLiuPub is possible to get the functionality to have parallel execution for detection models as similar to this PR? - #143. Would be very useful to get faster speeds especially on the Movidius VPU
Hi, @joelbudu
Do you need to input multiple data from several cameras? or only use one camera but want to execute the inference by several Movidius VPUs in parallel?
@LewisLiuPub I would like to use one camera but execute the inference by several Movidius VPUs in parallel, please.
@joelbudu Can you let me know more, which model(s) are you using in this case?
Object Detection -> YOLOV2
Object Segmentation -> MaskRCNN
To update the status for multiple pipeline supporting and yolo/maskrcnn supporting:
- Yolov2 is deprecated. Instead, Yolov5 is supported.
- instance segmentation (maskrcnn, deeplabv3) is supported on branch ros2 (enabled for ros2 foxy/galactic)
Thank you for working on this @LewisLiuPub