KGTN-ens: Few-Shot Image Classification with Knowledge Graph Ensembles
2022-11-06Code Available0· sign in to hype
Dominik Filipiak, Anna Fensel, Agata Filipowska
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/DominikFilipiak/KGTN-ensOfficialpytorch★ 4
Abstract
We propose KGTN-ens, a framework extending the recent Knowledge Graph Transfer Network (KGTN) in order to incorporate multiple knowledge graph embeddings at a small cost. We evaluate it with different combinations of embeddings in a few-shot image classification task. We also construct a new knowledge source - Wikidata embeddings - and evaluate it with KGTN and KGTN-ens. Our approach outperforms KGTN in terms of the top-5 accuracy on the ImageNet-FS dataset for the majority of tested settings.
Tasks
Benchmark Results
| Dataset | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| ImageNet-FS (10-shot, all) | KGTN-ens (ResNet-50, h+g, max) | Top-5 Accuracy (%) | 83.46 | — | Unverified |
| ImageNet-FS (10-shot, novel) | KGTN-ens (ResNet-50, h+g, max) | Top-5 Accuracy (%) | 82.56 | — | Unverified |
| ImageNet-FS (1-shot, all) | KGTN-ens (ResNet-50, h+g, max) | Top-5 Accuracy (%) | 68.58 | — | Unverified |
| ImageNet-FS (1-shot, novel) | KGTN-ens (ResNet-50, h+g, max) | Top-5 Accuracy (%) | 62.73 | — | Unverified |
| ImageNet-FS (2-shot, all) | KGTN-ens (ResNet-50, h+g, max) | Top-5 Accuracy (%) | 75.45 | — | Unverified |
| ImageNet-FS (2-shot, novel) | KGTN-ens (ResNet-50, h+g, max) | Top-5 Accuracy (%) | 71.48 | — | Unverified |
| ImageNet-FS (5-shot, all) | KGTN-ens (ResNet-50, h+g, max) | Top-5 Accuracy (%) | 81.12 | — | Unverified |
| ImageNet-FS (5-shot, novel) | KGTN-ens (ResNet-50, h+g, mean) | Top-5 Accuracy (%) | 78.9 | — | Unverified |