By chance, is it possible working in parallel with two Jetson Tx2?
In particular, my idea is testing Openpose ( GitHub - CMU-Perceptual-Computing-Lab/openpose: OpenPose: Real-time multi-person keypoint detection library for body, face, hands, and foot estimation ) benefiting from two tx2 working together to increase performance. At the moment Openpose works on only one Tx2 with low fps and low resolution.
Thank you!
Sure, it is most common to network multiple Jetson’s over Gigabit Ethernet as the interconnect. If you are using ROS, this is made more transparent via their transport layer. I am not sure what you would have to do to Openpose to explicitly integrate the intercommunication, or if it already supports that.
Thank you for the answer.
When I interconnect two Jetson tx2 over Gigabit Ethernet, am I actually using two resources to compute the same code in general at the same time as a cluster likewise NVIDIA SLI?
If it possible, let me know how to connect them (H/w or S/W), please.
If you are using a framework like MPI or ROS, they can automatically manage the distributed computing and assign tasks across the Jetson’s. Otherwise it will be up to your application to run certain code on certain Jetson’s, forming a coherent pipeline particular to your application.