Classifier returns [0.5, 0.5]

Hi im try to train a SSD-model,
but after faw epochs the classifer returns exactly [0.5, 0.5]…
what could be causing this problem?
i use for loss cross entropy and for optimizer optimizer = torch.optim.Adam(model.parameters(), lr=0.001)

To avoid any confusion, could you please specify what [0.5, 0.5] exactly represent here? Is it the proportion of samples in both categories, in case you have two classes?