-
### Description
The current SOTA paper for image retrieval was published in 2020. It's called "Combination of Multiple Global Descriptors for Image Retrieval," and you can find it [here](https://arxi…
-
-
I'm trying to use `metric-learn` to learn a mahalanobis-metric approximation for a precomputed distance metric (that's computed in an different way). Is this functionality supported by the library? I …
-
in the case of image classification,
we have a training set, with x inputs, and the loss is computed by whether our neural network was correctly able to predict the label or not, and accordingly ta…
-
### What
The embeddings we use is generating a metric space based on cosine similarity.
While computing nearest neighbours, we use euclidian distance.
Change this to use cosine similarity.
- Best don…
-
## 🚀 Feature
I would like to contribute to the pytorch (in python and c++) by adding the [ contrastive loss/siamese_margin_loss](http://yann.lecun.com/exdb/publis/pdf/hadsell-chopra-lecun-06.pdf) f…
-
@QianHaosheng @bugtig6351 @yuanpcr you can list all potential metrics for the `generate` task in this issue. For more details about the `generate` task, you can refer to issue #12 .
-
There's a list here: http://ntz-develop.blogspot.co.uk/2011/03/fuzzy-string-search.html
-
Assignment 2 aims to find misinformation on social network, i.e., identify profiles that
are mistakenly recorded as human/non-human profiles
-
Hi,
I want to know your reference(paper) about triplet loss.
Can you give me the paper you refer to? facenet??