Lets say that I've already trained a model on 500 images and saved it as a HDF5 file. Now I have 90 more images that I would like to train with. How exactly do I make it so that the this will improve the model, and not overwrite the model with 500 images?
Asked
Active
Viewed 82 times
0
-
Does this answer your question? [Loading a trained Keras model and continue training](https://stackoverflow.com/questions/42666046/loading-a-trained-keras-model-and-continue-training) – ilke444 Mar 03 '20 at 22:39
1 Answers
0
1) Since the dataset size is very minimal (500) you can retrain all the the data together.
Else
2) You can try transfer learning with a small learning rate on the 90 images. Keeping learning rate small will not drastically change the already trained weights.
More on transfer learning https://towardsdatascience.com/transfer-learning-from-pre-trained-models-f2393f124751
Pushpa
- 892
- 1
- 12
- 30