Tegra K1 Image Signal Processor

The tegra k1 has 2 ISPs as part of the camera subsystem. I’m bit new to tegra development but can anyone point me in the right direction as to what API or libraries would give access or allow me to write code that utilizes these two ISPs.

I didn’t find anything from TRM related to ISPs. What kind of things would you like to achieve by utilizing those directly instead of some later phase image processing API?

It’s a good question because nowadays many manufacturers of SoC are blocking access to their DSP advanced documentation (for advanced application) even to their SoCs customers when they are not big enough.
If Jetson TK1 can be used for prototyping, it would be really nice to know if the DSP is a blackbox or not.

What kind of this are you expecting to do with the ISP? I guess a lot of stuff can be done with CUDA even though ISP and CUDA (GPU) are designed for different goals.

If some one from Nvidia could just answer the question it would be nice, given the fact that there is only a very few questions asked, it shouldn’t be a problem to answer isn’t it ?

As to what application, i guess many would prefer to keep their reasons confidentials however in my own case, i’m looking for extensive scaling and frame rate conversion from camera to hdmi output. In such case, access to DSP documentation is a key point.

I would also be interested in using the ISP, either by connecting a cSI-2 camera directly, or using a USB industrial camera. In either case I would end up with raw data that I need to debayer and process.

So the question is what ressources are available in the ISP and how to program it ? I didn’t find anything in all the documentation I read so far. Any idea?

I believe image processing is performed on the ISP using the gstreamer interface.
Check out the nvgstcapture and nvgstplayer executables.

https://developer.nvidia.com/sites/default/files/akamai/mobile/files/L4T/gstomx_src.tbz2