Hi, there!

​

This is an example of image retrieval based on MNIST and fashion-MNIST!

After learning by using typical triplet loss (a type of metric loss), we checked the retrieval results for the query image.

​

Despite simple learning, the results show fairly accurate retrieval results.

We have also shown the embedding space for the learned network in various ways.

If you are interested in doing research or metric learning, you might want to refer to it.

​

https://github.com/kdhht2334/Triplet_loss_for_image_retrieval

​

Image retrieval result 1

Image retrieval result 2

Visualization of embedding space