for eval

[xuchen-dev]

very hopeful for your help：

i use your fastercnn.py as a moudle in my model , when train it is very good the result is good too

but when eval i just change the towersetup the istrain is false , but there is no output box mask

i try to find the reason, but i cant could you please teach me is there something else i need to do when i use the fasterrcnn to eval , so it can output box prob mask

it seem everthing is normal before the create heaeds

[xuchen-dev]

@Pandoro @pvoigtlaender @JonathonLuiten

[xuchen-dev]

it seem like there is somethings wrong in _create_fastrcnn_output

when istrain is faslse

when istrain true

[pvoigtlaender]

Hi xuchen,

note that there is no need to change the arguments to tower_setup. Instead, please change the config (create a separate config file for eval, or override using the command line).
The easiest should be to see how we do "forwarding and tracking" in the README.

Best,

Paul