Find images that produce feature vectors closest to the feature vector of test image.

[Hieu Do Trung]

As the title, how can I to find images that "produce feature vectors in the last hidden layer with the smallest Euclidean distance from the feature vector for the test image", just like this:

[Axel Angel]

It seems very easy with Python:

(1) load your network

(2) forward_all your dataset, blobs=['your_last_layer_name']

(3) the return value is a dict, take your_last_layer_name value which is a matrix (list of feature vectors)

(4) compute distances between your image feature and the rest of features, take top-k, where k=6 in this paper.

[Hieu Do Trung]

You are my hero :)
Will try it.
Thank you very much.

[Cuong Duc]

@Axel Angel: How to take the top-k?

Do I need to modify the accuracy layer in train_val.prototxt to take the top-k?

Thanks