Keras Retinanet - Custom Object Detector - Error on starting to detect on Nano

I trained and tested in my computer a Retinanet custom object detector.


Now, when i am trying it on the nano…

I am having a problem with it. I am using Retinanet (from the repo)

and below a general list of dependencies

JetPack 4.4 r32
Tensorflow 2.1.0
Keras 2.3.1

But you will see the installed packages requirements.txt (831 Bytes) .

Whei i run it there is an error about the attr of the version of tensorflow


I made a temp patch modifying the file at the line with error with


Next with the dependencies and this error solved i run my retinaned pre-trained model from my computer into the jetson to make inferences but got this error:

log_error.txt (14.4 KB)


Sorry for keeping you waiting.

This is a known issue and is fixed in our next release.
For now, please check this comment for workaround: