Is there a quick procedure to update this? It would be nice to get this working. Right now there’s an error with the tensorrt_model.py and the None type.
“The “object following” jupyter notebook example won’t work due to the required dependencies that had not been released by NVIDIA prior to the creation of the SparkFun JetBot image. These updates can be manually installed on your Jetson Nano with the JetPack 4.2.1 release.”
I have an open issue on [Nvidia’s GitHub repository. Looks like you have commented there as well.
I’ve tried rebuilding the latest release (as of a month ago) and I still haven’t had any luck getting the object following example to work. Unfortunately, I haven’t had time to dig into the issue further than the version requirements (I rebuilt the image from scratch at +15 times with the new release trying to bypass the version issue). I think this issue is in the TensorRT module from Nvidia, which is a little" deeper" than I can handle at the moment. I’ll try to reach out to Nvidia directly; I’ll update this forum if I hear back from Nvidia.
(Just a heads up, I’ll be out of town for 2 weeks w/o any internet or cell. Thanks for your patience.)](TensorRT version issue w/ Object Following · Issue #126 · NVIDIA-AI-IOT/jetbot · GitHub)
Yeah, it looks like a versioning issue with the neural net and tensorrt. It was my goal to present it at an AI robotics conference next week but I don’t really have time at this point. So I’ll probably play with it a bit more in between semesters and stick with the Coral Edge TPU for the conference 
I heard back from Nvidia.
The issue is actually with the engine built for that example; it is only compatible with an older version of TensorRT (used to build it). You would need to use the original Jetbot image that they released (I am still following up with where to get that; the current links appear to have the updated images). However, once you have that, follow the Re-Applying the Software Modifications of the [assembly guide to update the image for the hardware modifications.
For more details, they also followed up with a response to the [GitHub issue I filed. (*It seems like there is a way to [rebuild the engine for the newer TensorRT release.[/i])](GitHub - AastaNV/TRT_object_detection: Python sample for referencing object detection model with TensorRT)](TensorRT version issue w/ Object Following · Issue #126 · NVIDIA-AI-IOT/jetbot · GitHub)](Assembly Guide for SparkFun JetBot AI Kit - SparkFun Learn)
Someone found an old image… here is the forum post with details to get it working: https://forum.sparkfun.com/viewtopic.ph … 52#p213052