Continue training on dataset B only.

AI will slowly forget the dataset A, training should be faster than training from scratch on dataset B.

Continue training on merged dataset A+B (image below).

AI will be extended to work on both dataset A and dataset B. This is useful if you find out that in some cases your trained AI does not perform well.

When using the Continue training function, the whole AI is being retrained. If dataset A and dataset B are not similar, then this function is not recommended.

Figure 526. Continue training on merged dataset A+B. [Courtesy of Nikon Imaging Center, RIES at Hokkaido University]

The left image (red binary) is with trained only 3 frames, the right image (green binary) is with adding 2 trained images (together 5 frames) after continuous training. After the continuous training we can see improvement in detection, where AI is more sensitive to the cell type then was before continuous training.