Graph Relation Distillation for Efficient Biomedical Instance Segmentation

TL;DR

This paper introduces a graph relation distillation method for biomedical instance segmentation that efficiently transfers knowledge from complex teacher models to lightweight students, focusing on instance features, relations, and boundaries.

Contribution

It proposes novel intra- and inter-image graph distillation schemes that improve lightweight models' performance by capturing structured instance and boundary relations.

Findings

Student models achieve less than 1% parameters and 10% inference time of teachers.

The approach outperforms existing distillation methods on biomedical datasets.

Effective knowledge transfer of instance relations and boundaries is demonstrated.

Abstract

Instance-aware embeddings predicted by deep neural networks have revolutionized biomedical instance segmentation, but its resource requirements are substantial. Knowledge distillation offers a solution by transferring distilled knowledge from heavy teacher networks to lightweight yet high-performance student networks. However, existing knowledge distillation methods struggle to extract knowledge for distinguishing instances and overlook global relation information. To address these challenges, we propose a graph relation distillation approach for efficient biomedical instance segmentation, which considers three essential types of knowledge: instance-level features, instance relations, and pixel-level boundaries. We introduce two graph distillation schemes deployed at both the intra-image level and the inter-image level: instance graph distillation (IGD) and affinity graph distillation (AGD). IGD constructs a graph representing instance features and relations, transferring these two types of knowledge by enforcing instance graph consistency. AGD constructs an affinity graph representing pixel relations to capture structured knowledge of instance boundaries, transferring boundary-related knowledge by ensuring pixel affinity consistency. Experimental results on a number of biomedical datasets validate the effectiveness of our approach, enabling student models with less than $ 1\%$ parameters and less than $10\%$ inference time while achieving promising performance compared to teacher models.