I have build the jetson inference demos and started the detecnet-camera app, but as soon as I started the app, nano crashed. I tried it again but it crashed again. I am using dc barrel jack 5v 4amp power supply.
I changed the mode to 5W mode. In this mode, nano didnt crashed but in this mode I am getting a low fps. Although I am not able to measure the exact fps because the code didn’t output fps but the overall speed I am getting is very slow.
I also tried the demo imagenet-camera app but here I am getting good fps.
Below are my questions:
- Why the fps is a slow in detectnet and not in imagenet demo.?
- How can I also get approx FPS from the demo.?
- Is there any demo where multiple models are used, just to check the overall performance of the nano.?
1. The complexity are different.
The detection is much more complicated than the image classification.
2. The fps is shown on the Window titles.
3. Deepstream sample does although it is not available yet.
Please wait for our release announcement, which supposed in the Q2 2019.
Here is a related blog for your reference:
Multi-Stream Video Analytics
Thanks for your response AastaLLL.
Do you have any idea why nano crashed during first time with default 10W mode.? It should not crash as I am using 5v 4amp supply.?
Hi abhinavrawat, which DC barrel jack power supply adapter are you using? Is it one from this thread?
If the voltage the adapter supplies dips below 4.75V under load, then the Nano can brownout and shutdown.