I hav a project that would use two 160 FOV cameras producing a 180 live stream video on youtube.
Is it feasible in performance?
What modules to use more than gstreamer ?
How to dewarp, and stitch into one stream?
/Fredrik
Hi fredrik,
Welcome to Jetson & Embedded Systems forum.
I don’t think there will be performance issue, but you may need to provide more details of the requirement of your use case, the other developers can share their thoughts. you also can refer to Jetson Projects forum to gain some ideas to develop your project.
OK, I’ll give an update on the project here, and you could move the topic if needed.
The harware setup for the project, as it looks now:
- Nvidia Jetson Xavier NX
- 2 x IMX477 (4k) camera sensors, probably the Raspberry HQ CAM
- Wide angle lenses with a HFOV of > 100 deg
So the project is in different stages:
Stage 1 - Create 4kx30fps 180 degree live streaming
- Cropping the two cameras
- Dewarping
- Aligning the images
- Stitching the two into one stream
- Encode into 180 degree video stream
- Live stream on Youtube
The following stages will apply overlaying and AI/ML functionality
Q1: If I’m using a lens with 130 deg HFOV, how much is worth cropping or should I dewarp the full image. Are there library functions for dewarping?
Q2: What encoding is required for running e.g. MPEG-4 4k in 180 degree format?
More questions to come
/Fredrik
hello fredrik.mattsson,
FYI,
you may refer to Jetson Xavier NX Software Features to evaluate your use-case.
for example,
CSI and USB Camera Features indicate that’s able to stream dual 4K (3840x2160) sensor up-to 60-fps.
you may also check Video Encoders session for the details of different encode methods.
please also check NVIDIA DeepStream SDK, there’s dewarp feature in DeepStream-4.0, which is GPU accelerated.
thanks
1 Like