peachthief

If you need a larger size you are probably best to provide it in patches and then reconstruct it. There are some pre-existing python modules to do so (https://github.com/adamrehn/slidingwindow for...

If you use the VOC2012 development kit it won't be arrange as needed for the training, as it has all the jpegimages together without a split between train and test....

The RTX cards have problems with older cudnn versions I've found, I have it running using tensorflow-gpu=1.13.2 cudatoolkit=10.0.130 cudnn=7.6.0 No changes to the code necessary

It is general issue with keras/tboard that you cannot get histograms with a validation_generator, changing histogram_freq=0 stops it trying to produce them.

@mueller91 you can pass validation data separately to generate the histograms I think in a separate step, but yeah it isn't whats needed really.