Loss values are not converging to a lower value.

[Sanath1998]

Hi @CoinCheung

Thanks for this wonderful real time segmentation repo. Actually I am trying to train with the custom dataset but seems like loss values is not converging to lower value. It is converged to a value of 2.8 or so. Is this the case with you as well for custom trained dataset.

If yes, what additional changes we need to do with respect to parameters/hypermeters? Please let me know which would be very beneficial.