I downloaded and install Jetpack 3.2 on my TX2 yesterday. Today I was trying the 2 days to a demo and everything compiled Ok and the two static image tests worked. But when I started ./imagenet-camera the video is…well…not good. There is a lot of noise and it has a very green tint to it. As such the image classification does not work very well.
So, is my camera goofy? Does the 2 days to a demo now work with JetPack 3.2? Is there something else I need to set up?
Any and all suggestion welcome!
Thanks,
Doug
Hi,
Could you share more information about your camera device?
The default setting of Jetson_inference is for CSI camera.
If you are using a USB camera, please remember to update the configuration into your use case.
Document: [url]https://github.com/dusty-nv/jetson-inference#running-the-live-camera-recognition-demo[/url]
Source: [url]https://github.com/dusty-nv/jetson-inference/blob/master/imagenet-camera/imagenet-camera.cpp#L37[/url]
Thanks.
Hi,
Oh good grief! It would help if I stated what the camera was/is. Sorry. I am using the built in camera that comes on the TX2, the onboard CSI camera. I have a SDI-USB video converter I would like to try but at the moment I am just using what was already available.
Thanks,
Doug
Hi,
Could you test if your camera can be opened via this GStreamer command in cmd first?
[url]https://github.com/dusty-nv/jetson-inference/blob/master/imagenet-camera/imagenet-camera.cpp#L37[/url]
Thanks.
AastaLLL,
I ran the imagenet-camera from the jetson-inference project and I still have a very green tint to everything. I have a snap shot and will try to upload.
![]()
Thanks,
Doug


does it seem like green in daylight?
if you run nvgstcapture?
by default camera have some protector surface over it ; if you did not remove it it will impact image color
Andrey1984,
I just managed to get a jpg up, I haven’t used nvgstcapture but I will figure it out and see what it shows me.
Thanks!
Doug
try
nvgstcapture-1.0
in my case in night time it produces obscure image, but in daylight is it fine
Andrey1984,
Well I am a complete idiot it seems as there was this lovely green lens protector on the camera. Yes, I feel stupid!
I am rebuilding jetson-inference now and I suspect it will work much, much better!
[edit]
Yes the video is much better, but the classification leaves quite a bit to be desired! However, that can be worked on.
Thanks for your help!
Doug
The lens protector tends to get lost, unfortunately, I have lost mine ages ago.
The next thing that tends to get lost is the camera itself… fortunately every time it detached I managed to find it and attach back yet. Not sure how many detach-attach times the connector will survive, though. And where to order a spare camera in case of final lost.