How many features i need select from a set ranked image feature space?

I was trained a pretrained DCNN on my image dataset, extracted the features form the last fully coneected layer and I applied filter type feature slection method. Which produces the ranks and scores for the feature space.How many features i need to select to improve the performance of the model.

Answer 1

[yanqi liu]

yes，sir，may be use activations to get the vector，such as

activations(net,data,layer,'OutputAs','rows');