Hyperbolic Image Embeddings
2019-04-03CVPR 2020Code Available1· sign in to hype
Valentin Khrulkov, Leyla Mirvakhabova, Evgeniya Ustinova, Ivan Oseledets, Victor Lempitsky
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/KhrulkovV/hyperbolic-image-embeddingsOfficialIn paperpytorch★ 0
- github.com/leymir/hyperbolic-image-embeddingsOfficialIn paperpytorch★ 0
- github.com/nalexai/hyperlibtf★ 147
Abstract
Computer vision tasks such as image classification, image retrieval and few-shot learning are currently dominated by Euclidean and spherical embeddings, so that the final decisions about class belongings or the degree of similarity are made using linear hyperplanes, Euclidean distances, or spherical geodesic distances (cosine similarity). In this work, we demonstrate that in many practical scenarios hyperbolic embeddings provide a better alternative.
Tasks
Benchmark Results
| Dataset | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| CUB 200 5-way 1-shot | Hyperbolic ProtoNet | Accuracy | 60.52 | — | Unverified |
| CUB 200 5-way 5-shot | Hyperbolic ProtoNet | Accuracy | 72.22 | — | Unverified |
| Mini-Imagenet 5-way (1-shot) | Hyperbolic ProtoNet | Accuracy | 51.57 | — | Unverified |
| Mini-Imagenet 5-way (5-shot) | Hyperbolic ProtoNet | Accuracy | 66.27 | — | Unverified |
| OMNIGLOT - 1-Shot, 20-way | Hyperbolic ProtoNet | Accuracy | 95.9 | — | Unverified |
| OMNIGLOT - 1-Shot, 5-way | Hyperbolic ProtoNet | Accuracy | 99 | — | Unverified |
| OMNIGLOT - 5-Shot, 20-way | Hyperbolic ProtoNet | Accuracy | 98.15 | — | Unverified |
| OMNIGLOT - 5-Shot, 5-way | Hyperbolic ProtoNet | Accuracy | 99.4 | — | Unverified |