ShapeEmbed: a self-supervised learning framework for 2D contour quantification

2025-07-01

AI Summary

This work addresses the challenge of achieving geometric invariance and representation generalizability in quantitative 2D object contour shape analysis. We propose a self-supervised contrastive learning framework for shape descriptor learning. Methodologically, we encode contour structure via Euclidean distance matrices and design a deep embedding mechanism invariant to translation, scaling, rotation, reflection, and point indexing—without relying on explicit data augmentation or handcrafted priors. Compared with conventional shape descriptors and existing autoencoder-based approaches, our framework preserves strict geometric invariance while significantly improving discriminability and robustness of shape representations. Extensive experiments on multi-source natural and biological image datasets demonstrate that our method achieves state-of-the-art performance in shape classification tasks, validating its strong generalization capability and practical applicability.

Technology Category

• Computer Vision: Representation Learning for Vision
• Machine Learning: Unsupervised & Self-Supervised Learning
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

The shape of objects is an important source of visual information in a wide range of applications. One of the core challenges of shape quantification is to ensure that the extracted measurements remain invariant to transformations that preserve an object's intrinsic geometry, such as changing its size, orientation, and position in the image. In this work, we introduce ShapeEmbed, a self-supervised representation learning framework designed to encode the contour of objects in 2D images, represented as a Euclidean distance matrix, into a shape descriptor that is invariant to translation, scaling, rotation, reflection, and point indexing. Our approach overcomes the limitations of traditional shape descriptors while improving upon existing state-of-the-art autoencoder-based approaches. We demonstrate that the descriptors learned by our framework outperform their competitors in shape classification tasks on natural and biological images. We envision our approach to be of particular relevance to biological imaging applications.

Problem

• Quantify 2D object contours invariantly to geometric transformations
• Overcome limitations of traditional and autoencoder-based shape descriptors
• Improve shape classification performance in natural and biological images

Innovation

• Self-supervised learning for 2D contour quantification
• Euclidean distance matrix encoding for shape descriptors
• Invariant to translation, scaling, rotation, and reflection