A multi-task U-net for segmentation with lazy labels

Abstract : The need for labour intensive pixel-wise annotation is a major limitation of many fully supervised learning methods for image segmentation. In this paper, we propose a deep convolutional neural network for multi-class segmentation that circumvents this problem by being trainable on coarse data labels combined with only a very small number of images with pixel-wise annotations. We call this new labelling strategy 'lazy' labels. Image segmentation is then stratified into three connected tasks: rough detection of class instances, separation of wrongly connected objects without a clear boundary, and pixel-wise segmentation to find the accurate boundaries of each object. These problems are integrated into a multitask learning framework and the model is trained end-to-end in a semi-supervised fashion. The method is applied on a dataset of food microscopy images. We show that the model gives accurate segmentation results even if exact boundary labels are missing for a majority of the annotated data. This allows more flexibility and efficiency for training deep neural networks that are data hungry in a practical setting where manual annotation is expensive, by collecting more lazy (rough) annotations than precisely segmented images.
Document type :
Preprints, Working Papers, ...
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-02170180
Contributor : Nicolas Papadakis <>
Submitted on : Monday, July 1, 2019 - 5:04:50 PM
Last modification on : Tuesday, July 2, 2019 - 1:47:24 AM

Links full text

Identifiers

  • HAL Id : hal-02170180, version 1
  • ARXIV : 1906.12177

Collections

Citation

Rihuan Ke, Aurélie Bugeau, Nicolas Papadakis, Peter Schuetz, Carola-Bibiane Schönlieb. A multi-task U-net for segmentation with lazy labels. 2019. ⟨hal-02170180⟩

Share

Metrics

Record views

7