Instance segmentation with the number of clusters incorporated in embedding learning

TL;DR

This paper introduces FCRNet, a one-stage instance segmentation method that incorporates the number of clusters into the embedding space, reducing errors and improving performance on nucleus images.

Contribution

The paper proposes embedding prior clustering information into the learning framework to enable one-stage instance segmentation, simplifying the process and enhancing accuracy.

Findings

FCRNet outperforms existing methods on the BBBC006 dataset.

Embedding cluster count into the model improves segmentation accuracy.

One-stage approach reduces segmentation errors compared to multi-stage methods.

Abstract

Semantic and instance segmentation algorithms are two general yet distinct image segmentation solutions powered by Convolution Neural Network. While semantic segmentation benefits extensively from the end-to-end training strategy, instance segmentation is frequently framed as a multi-stage task, supported by learning-based discrimination and post-process clustering. Independent optimizations on substages instigate the accumulation of segmentation errors. In this work, we propose to embed prior clustering information into an embedding learning framework FCRNet, stimulating the one-stage instance segmentation. FCRNet relieves the complexity of post process by incorporating the number of clustering groups into the embedding space. The superior performance of FCRNet is verified and compared with other methods on the nucleus dataset BBBC006.