Questions tagged [triplet-loss]
Triplet losses are defined in terms of the contrast between three inputs. Each of the inputs has an associated class label, and the goal is to map all inputs of the same class to the same point, while all inputs from other classes are mapped to different points some distance away. It's called a triplet because the loss is computed using an anchor, a sample belonging to the same class as the anchor, and a sample belonging to a different class.
18
questions
2
votes
0
answers
95
views
Euclidean and geodesic distance have different gradients. Does mixing the two concepts impair triplet learning?
The triplet loss is defined by Florian Schroff, Dmitry Kalenichenko, James Philbin in "FaceNet: A Unified Embedding for Face Recognition and Clustering" as
$$
\mathcal L = \sum_\mathcal T \...
- 87.5k
1
vote
1
answer
2k
views
How to use cosine similarity within triplet loss
The triplet loss is defined as follows:
$$
L(A, P, N) = max(‖f(A) - f(P)‖² - ‖f(A) - f(N)‖² + margin, 0)
$$
where $A$=anchor, $P$=positive, and $N$=negative are the data samples in the loss, and $...
- 633
1
vote
0
answers
106
views
How to use a Siamese network at test time? [closed]
I am trying to understand Siamese networks, and understand how to train them.
Once I have a trained network, I want to know if a new image is close or far to other images in the train set, and fail to ...
- 453
2
votes
0
answers
855
views
Why does cotrastive loss and triplet loss have the margin element in them?
Trying to learn Siamese networks for ranking tasks from here, I find it hard to understand why triplet loss and contrastive losses have the "margin" element in them.
The contrastive loss
$L(...
- 453
1
vote
0
answers
209
views
Why does triplet loss outperform contrastive loss?
Trying to learn Siamese networks for ranking tasks from here, I find it hard to understand why triplet loss was ever introduced at all, theoretically. I understand it works better in practice, but ...
- 453
1
vote
0
answers
274
views
Resnet model does not train with triplet loss, while VGG16 is able to train, why?
I am trying to do a transfer learning with ResNet50V2 model using triplet loss function. I have kept Include_top = False, input shape = (160,160,3) with Imagenet weights. The last 3 layers of my model ...
1
vote
0
answers
266
views
Why are we interested in gradient with respect to input?
I am learning about sampling methods for Deep Embedding Learning. I was reading an article named: "Sampling Matters in Deep Embedding Learning" (https://arxiv.org/abs/1706.07567).
In the ...
- 11
7
votes
1
answer
10k
views
In training a triplet network, I first have a solid drop in loss, but eventually the loss slowly but consistently increases. What could cause this?
I haven't even finished 1 epoch, so I don't think it could any sort of overfitting. I am training on a very large amount of data (27 gb of text) so it'll still be a while before I even reach one epoch....
- 989
0
votes
0
answers
25
views
Is there any algorithms/models to generate embedings of sequential data (other than RNN)?
I know that RNN can be used for such task. For instance facenet used rnn with triplet loss. But maybe there are some less sophisticated alternatives to try first?
- 103
3
votes
1
answer
80
views
Learning useful semantic representations of data
Training a neural network on its final task (e.g. classification) right from the beginning is not always the best way to go. I'd like to make a short list of recognized methods of motivating a NN to ...
- 205
0
votes
0
answers
292
views
Replacement for angular distance metric
I am looking for a distance metric that could be used instead of cosine/angular distance for high dimensional data. Metric that is limited the same way as cosine/angular distance is would be great.
...
- 1
2
votes
1
answer
126
views
Maximizing AUC based on point cloud distance
Let $V$ be an $n$ dimensional space with sets of positive class vectors $P$ and negative class vectors $N$. The task is to find a vector $x$ such that AUC is maximized, based on ranking generated by ...
- 13.7k
2
votes
2
answers
1k
views
Why the Triplet Loss function distincts anchor and positive?
I read the paper and I understand that anchoring one image and select corresponding semi-hard positives and negatives is an efficient way of generating samples.
However, I don't understand why the ...
- 121
3
votes
1
answer
760
views
Facenet: Using Ensembles of Face Embedding Sets
The Facenet is a deep learning model for facial recognition. It is trained for extracting features, that is to represent the image by a fixed length vector called embedding. After training, for each ...
- 131
1
vote
2
answers
2k
views
Does triplet loss help for document similarity search?
I build CNN network over documents with triplet loss. And compare documents with cosine similarity. It does really find similar docs and catches interesting dependencies. But simple tfidf model does ...
- 139
1
vote
1
answer
526
views
Overcome underfitting on train data using CNN architecture
I use 2 layer CNN network for NLP task with triplet loss with margin 0.2. The task is to learn document embeddings to find similar docs.
My architecture is similar to this https://arxiv.org/abs/1406....
- 139
5
votes
0
answers
1k
views
Triplet Deep Learning Embedding Loss Functions
Triplet embeddings consist of mapping a group of images to an embedding space, such that images deemed more similar to each other end up closer together. The "triplet" comes from training, ...
- 13.7k
6
votes
1
answer
2k
views
How does FaceNet (Google's facerecognition) handles a new image?
I am currently researching in the facerecognition field.
And I can not understand how the facenet algorithm handels a new image
They use an euclidean space for image representation. Which means that ...
- 163