Can Euclid help with this

[RTasa]

I want to combine Tensorflow inference data with distances to the object in an outdoor environment.

Basically a way to delineate people in the vicinity of a moving vehicle from other equipment which also moves.

In my mind I was thinking LiDar might be helpful for this but a rotating unit cost is too high.

Even if I could bring the cost down I am not sure how to pair it up with actual inference data.

I have thought about stereo imaging but from what I understand the processing

power is so great that its not used.

So in an outdoor environment can the Euclid provide distance information about 200 feet out?

[MartyG]

An excellent fit for your project may be the new powerful RealSense cameras D415 and D435 that are being released next month. They were designed with autonomous vehicles and drones in mind and have a vision processing chip with real-time processing, as well as 90 FPS frame rate and long range sensing.

https://software.intel.com/en-us/realsense/400-series Intel® RealSense™ Depth Cameras D400-Series | Intel® Software

[RTasa]

It sounds promising. Can you point to any information on these cameras?

[MartyG]

There is currently not much more information available on these models other than what is on the page I linked to. More information will be available when the cameras are released, and you should be able to see further specification details on the cameras' listings in the Intel Click online store once they appear there.

https://click.intel.com/realsense.html Intel® RealSense™ Developer Kits

I can add that they are not compatible with the current RealSense SDK application development software because they use a new software stack, and so will use a new cross-platform open source SDK.

You may also get some useful insight from a YouTube video that discusses the new model when it was known by the codename R400.

https://www.youtube.com/watch?v=pvXJSn22ujU&t=6s Intel RealSense 400 for self-driving cars, autonomous drones, Robots, for indoor and outdoor use - YouTube

[RTasa]

Ahhh now I get it. I have seen YouTube video but had no idea where they were in development.

The D# did not mean R400 to me. This is great news and pretty good timing. Can't wait for the details.

I did some research into the Movidius stuff but was disappointed that it did not support TensorFlow Objects or YOLO

or R-CNN etc. I hope Intel can make a co-processor that does that.