Darknet has the option to resize the network during training and adapt to different resolutions during training.
Do you know whether this is possible in tensorflow? I thought of recompiling the network, but not sure whether for example the current optimizer parameters can be kept.
Hello love your repo, it's super clean!
Darknet has the option to resize the network during training and adapt to different resolutions during training. Do you know whether this is possible in tensorflow? I thought of recompiling the network, but not sure whether for example the current optimizer parameters can be kept.
I would also implement it if you give me a hint.