Is the Jetson TX2 the right equipment?
I am working on a robotics project that requires very fast object recognition (ideally 50fps+) with very low latency on the video input (sub 10ms ideally) and of course minimal power usage. When the project is done it will need to recognize and track a person but for now, I can just use anything as a stand-in for a proof of concept. The robot needs to be able to react extremely quickly to real-world situations. Right now I am using the CMU pixy camera, at this phase, I only need it to recognize colored objects so the pixy works, but this is holding me back on lens/resolution. I am working on the next version and am thinking the TX2 is possibly the best option, any input? Does this sound like a good job for the TX2? Thanks!
I am working on a robotics project that requires very fast object recognition (ideally 50fps+) with very low latency on the video input (sub 10ms ideally) and of course minimal power usage. When the project is done it will need to recognize and track a person but for now, I can just use anything as a stand-in for a proof of concept.

The robot needs to be able to react extremely quickly to real-world situations. Right now I am using the CMU pixy camera, at this phase, I only need it to recognize colored objects so the pixy works, but this is holding me back on lens/resolution.

I am working on the next version and am thinking the TX2 is possibly the best option, any input? Does this sound like a good job for the TX2?

Thanks!

#1
Posted 01/13/2018 07:18 AM   
Hi brandonlima, please refer to these benchmarks showing image recognition latency down to 5.6ms (GoogLeNet @ 179 FPS) with TX2 and TensorRT: [url]https://devblogs.nvidia.com/parallelforall/jetpack-doubles-jetson-inference-perf/[/url] For TX2 power benchmarks, see this article: [url]https://devblogs.nvidia.com/parallelforall/jetson-tx2-delivers-twice-intelligence-edge/[/url] To deploy a more compact TX2, see many of the miniaturized carriers and enclosures available: [url]https://elinux.org/Jetson_TX2#Ecosystem_Products[/url]
Answer Accepted by Forum Admin
Hi brandonlima, please refer to these benchmarks showing image recognition latency down to 5.6ms (GoogLeNet @ 179 FPS) with TX2 and TensorRT:

https://devblogs.nvidia.com/parallelforall/jetpack-doubles-jetson-inference-perf/

For TX2 power benchmarks, see this article: https://devblogs.nvidia.com/parallelforall/jetson-tx2-delivers-twice-intelligence-edge/

To deploy a more compact TX2, see many of the miniaturized carriers and enclosures available: https://elinux.org/Jetson_TX2#Ecosystem_Products
#2
Posted 01/14/2018 11:09 PM   
Thanks!
Thanks!

#3
Posted 01/21/2018 07:35 AM   
Scroll To Top

Add Reply