Dung-Han-Lee
commented
2 years ago can this code take the input from the lidar directly?
You will need to first project your pointcloud into 2D images, after all, that's what the model been trained for.
what I mean is that can I use this code for real-time application?
That's a totally different issue. It depends on what you meant by "realtime". For any robotic application, using python to run the code on cpu? I don't think anything on the GitHub will suit your requirements. If you use GPU, then you should be able to achieve realtime. If not, i.e you are intending to run this on cpu, then you'll need to look at opencv, they support some neural networks in c++. You will need to first train your model in pytorch, tensor flow etc, get the weight file, and apply it in opencv. But even so, on Intel i7, I imagine you'd achieve something between 5hz to 10 hz. If that's still too slow, then you need to rethink about the whole thing. Pointcloud processing is computationally intense in general.
probably too slow.
On Fri, Jul 15, 2022, 3:33 AM urbansound8K @.***> wrote:
can this code take the input from the lidar directly?
what I mean is that can I use this code for real-time application?
— Reply to this email directly, view it on GitHub https://github.com/Dung-Han-Lee/RIUNET-Row-Detection-with-LIDAR-range-image/issues/1, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGY7SKV6OLXH2RWEZF34D6LVUEH5RANCNFSM53UVODCA . You are receiving this because you are subscribed to this thread.Message ID: <Dung-Han-Lee/RIUNET-Row-Detection-with-LIDAR-range-image/issues/1@ github.com>
can this code take the input from the lidar directly?
what I mean is that can I use this code for real-time application?