It looks that the overfitting is not occuring in your case since the validation performance still keep improving.
Overfitting occurs when training performance is still improving while validation performance becomes worse.
For your case, both training and validation performances are still improving (error reducing) so you can keep training the model for further performance improvement.
The "best" (circle on green curve) shows that the latest model is the best. Thus there is still room for further improvement.
