Docker version 20.09 caffe inclusion and working object detection example.

[Graeme Murphy]

I installed the docker version Digits 20.09

It installed correctly and it ran up ok..

I did notice that in the 'Info' menu 'Caffe version: unavailable' and 'Caffe flavor: Unavailable'

This version of the docker file does not seem to include caffe ??

Also....

I ran the Classification example.

https://github.com/NVIDIA/DIGITS/tree/master/examples/classification

This ran Ok. and seemed to be all Ok...

I then ran the Object Detection example: 

https://github.com/nvidia/digits/tree/master/examples/object-detection

This created the data set however however I have not been able to build the model, firstly because the example relies on you having caffe available. When I tried to run it using 'Tensorflow' I have not been able to build the model. I have tried every combination possible.  If I try and use a 'custom network' when I click on visualize it shows 

#####################################

500: INTERNAL SERVER ERROR

TypeError a bytes-like object is required, not 'str'

#####################################

I have my own dataset I am trying to run... So I wanted to prove the Digits installation was running correctly as I have had trouble with my own dataset!

I am running the Docker file on a PC with only 3G of video ram, so its a bit limited..

Does anyone know WHICH set of instructions to follow to get caffe working on the digits docker file 20.09... and/or does anyone know of an example for 'object detection' which actually runs on the digits docker file 20.09 ...or am I missing something obvious!

My objective is I need to have a model for a device which runs SSD mobilenet ... So I am trying to create a model for that device ... or should I give up on Digits and use another builder! any suggestions are welcome at this stage...