Improved Few-Shot Visual Classification

By Peyman Bateni, Raghav Goyal, Vaden Masrani, Frank D. Wood, Leonid Sigal. 2019

272 citations recorded by the source corpus.

Few-shot learning is a fundamental task in computer vision that carries the promise of alleviating the need for exhaustively labeled data. Most few-shot learning approaches to date have focused on progressively more complex neural feature extractors and classifier adaptation strategies, and the refinement of the task definition itself. In this paper, we explore the hypothesis that a simple class-covariance-based distance metric, namely the Mahalanobis distance, adopted into a state of the art few-shot learning approach (CNAPS) can, in and of itself, lead to a significant performance improvement. We also discover that it is possible to learn adaptive feature extractors that allow useful estimation of the high dimensional feature covariances required by this metric from surprisingly few samples. The result of our work is a new ``Simple CNAPS'' architecture which has up to 9.2% fewer trainable parameters than CNAPS and performs up to 6.1% better than state of the art on the standard few-shot image classification benchmark dataset.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Memory Efficient Meta-Learning with Large Images (2021, methodology): “Experiments We meta-train ProtoNets [3], CNAPs [4] and Simple CNAPs [5] with LITE on tasks composed of large (224× 224) images.”
A Comprehensive Survey of Few-shot Learning: Evolution, Applications, Challenges, and Opportunities (2022, methodology): “In the past five years, we systematically combed and summarized FSL in the field of computer vision [161] and divided tasks into image classification, object detection, semantic segmentation, and instance segmentation.”
Improving Task Adaptation for Cross-domain Few-shot Learning (2021, methodology): “As in [2, 11, 22], we build our method on ResNet-18 [13] backbone in all experiments.”
Learning a Universal Template for Few-shot Dataset Generalization (2021, methodology): “…> . work is shifting towards a more challenging instance of few-shot classiﬁcation across datasets (Chen et al., 2019; Tseng et al., 2020; Triantaﬁllou et al., 2020; Requeima et al., 2019; Bateni et al., 2020; Bronskill et al., 2020; Saikia et al., 2020; Dvornik et al., 2020; Liu et al., 2021).”
FiT: Parameter Efficient Few-shot Transfer Learning for Personalized and Federated Image Classification (2022, methodology): “The results for all models (FineTuner (Yosinski et al., 2014), Simple CNAPs (Bateni et al., 2020) and ProtoNets (Snell et al., 2017)) are from (Bronskill et al., 2021).”
A Universal Representation Transformer Layer for Few-Shot Image Classification (2020, methodology): “[1] with Simple CNAPS, leading to one of the current best methods on Meta-Dataset.”
Universal Representation Learning from Multiple Domains for Few-shot Classification (2021, methodology): “In addition, similar to Simple CNAPS [2], we map our features to a task-specific space before applying the nearest neighbor classifier but we learn the parameters of this mapping from each support set.”