Training deep neural networks take a lot of time and computation. In this paper, we have discussed how we can reduce the training time for deep learning models if we have already trained a model for a similar dataset. The basic logic here is that for similar dataset the features stored in the deep neural net are similar and the only difference comes for the classification layers of deep neural so instead of training the whole net we just train the last layers for classifying the data and use the trained model weights on the rest of the layers; this method saves a lot of time. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd 2021.