Graph-based Interpolation of Feature Vectors for Accurate Few-Shot Classification

27 Jan 2020  ·  Yuqing Hu, Vincent Gripon, Stéphane Pateux ·

In few-shot classification, the aim is to learn models able to discriminate classes using only a small number of labeled examples. In this context, works have proposed to introduce Graph Neural Networks (GNNs) aiming at exploiting the information contained in other samples treated concurrently, what is commonly referred to as the transductive setting in the literature. These GNNs are trained all together with a backbone feature extractor. In this paper, we propose a new method that relies on graphs only to interpolate feature vectors instead, resulting in a transductive learning setting with no additional parameters to train. Our proposed method thus exploits two levels of information: a) transfer features obtained on generic datasets, b) transductive information obtained from other samples to be classified. Using standard few-shot vision classification datasets, we demonstrate its ability to bring significant gains compared to other works.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Benchmark
Few-Shot Image Classification CUB 200 5-way 1-shot Transfer+SGC Accuracy 88.35% # 12
Few-Shot Image Classification CUB 200 5-way 5-shot Transfer+SGC Accuracy 92.14 # 11
Few-Shot Image Classification Mini-ImageNet - 1-Shot Learning Transfer+SGC Accuracy 76.47% # 5

Methods


No methods listed for this paper. Add relevant methods here