A prediction box problem generated by the model #35

Thank you for your contribution. I would like to ask why the length and width of the prediction box generated by the model I trained are close to 1. The output references are sized like this [2.3472e-01, 5.9792e-01, 9.9979e-01, 9.9223e-01],
[7.3743e-01, 2.3905e-01, 9.3747e-01, 9.8086e-01]。 Is it the reason for my dataset?