FAAGC: Feature Augmentation on Adaptive Geodesic Curve Based on the shape space theory

2025-01-25

AI Summary

To address the sparsity and insufficient discriminability of feature representations in few-shot learning, this paper introduces shape space theory into feature enhancement for the first time. Specifically, deep features are embedded into the pre-shape space, leveraging the geometric prior that shape-equivalent objects lie on great circles of the unit sphere. For each class, a differentiable, adaptive geodesic curve is constructed on the Riemannian manifold, and semantically consistent novel features are generated via spherical interpolation along the curve. The method comprises four key components: pre-shape space embedding, geodesic curve modeling on the Riemannian manifold, adaptive parametric sampling, and geometric regularization. Evaluated on multiple few-shot benchmarks, the approach achieves consistent classification accuracy gains of 3.2–7.8%. Notably, it demonstrates strong generalization across diverse backbone architectures—including CNNs, Vision Transformers (ViTs), and CLIP—without architectural modifications.

Technology Category

• Machine Learning: Learning with Manifolds
• Computer Vision: Representation Learning for Vision
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Deep learning models have been widely applied across various domains and industries. However, many fields still face challenges due to limited and insufficient data. This paper proposes a Feature Augmentation on Adaptive Geodesic Curve (FAAGC) method in the pre-shape space to increase data. In the pre-shape space, objects with identical shapes lie on a great circle. Thus, we project deep model representations into the pre-shape space and construct a geodesic curve, i.e., an arc of a great circle, for each class. Feature augmentation is then performed by sampling along these geodesic paths. Extensive experiments demonstrate that FAAGC improves classification accuracy under data-scarce conditions and generalizes well across various feature types.

Problem

• Deep Learning
• Small Dataset
• Model Accuracy

Innovation

• FAAGC
• pre-shape space
• data augmentation