Semantic Segmentation via Multi-task, Multi-domain Learning

Fourure, Damien; Emonet, Rémi; Fromont, Élisa; Muselet, Damien; Trémeau, Alain; Wolf, Christian

doi:10.1007/978-3-319-49055-7_30

Cited by 6 publications

(9 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This paper extends [38]. We introduced another method (called Joint train-ing with shared context) that can use the correlations between the labelsets learned by the network to improve the accuracy.…”

mentioning

confidence: 94%

Multi-task, multi-domain learning: Application to semantic segmentation and pose regression

Fourure¹,

Emonet²,

Fromont³

et al. 2017

Neurocomputing

Self Cite

View full text Add to dashboard Cite

We present an approach that leverages multiple datasets annotated for different tasks (e.g., classification with different labelsets) to improve the predictive accuracy on each individual dataset. Domain adaptation techniques can correct dataset bias but they are not applicable when the tasks differ, and they need to be complemented to handle multi-task settings. We propose a new selective loss function that can be integrated into deep neural networks to exploit training data coming from multiple datasets annotated for related but possibly different label sets. We show that the gradient-reversal approach for domain adaptation can be used in this setup to additionally handle domain shifts. We also propose an auto-context approach that further captures existing correlations across tasks. Thorough experiments on two types of applications (semantic segmentation and hand pose estimation) show the relevance of our approach in different contexts.

show abstract

mentioning

confidence: 94%

Multi-task, multi-domain learning: Application to semantic segmentation and pose regression

Fourure¹,

Emonet²,

Fromont³

et al. 2017

Neurocomputing

Self Cite

View full text Add to dashboard Cite

show abstract

“…To address the similar problem of limited annotated data for semantic scene labeling in natural images, Fourure et al proposed to train a single network over the union of multiple datasets, in order to leverage a greater amount of training data [7]. Their model is optimized in a multi-task and multi-domain framework, in which each dataset is defined by its own task (segmentation label-set) and domain (data distribution).…”

Section: Introductionmentioning

confidence: 99%

“…Their model is optimized in a multi-task and multi-domain framework, in which each dataset is defined by its own task (segmentation label-set) and domain (data distribution). Hence, this approach is more generic than traditional domain adaptation techniques which usually focus on domains containing the same set of objects [7]. Following this, Rebuffi et al [19,20] proposed to employ a model with agnostic filters, as visual primitives may be shared across tasks and domains, and dataset-specific layers which allow task and domain specialization.…”

Section: Introductionmentioning

confidence: 99%

Multi-Task, Multi-Domain Deep Segmentation with Shared Representations and Contrastive Regularization for Sparse Pediatric Datasets

Boutillon,

Conze,

Pons

et al. 2021

Preprint

View full text Add to dashboard Cite

Automatic segmentation of magnetic resonance (MR) images is crucial for morphological evaluation of the pediatric musculoskeletal system in clinical practice. However, the accuracy and generalization performance of individual segmentation models are limited due to the restricted amount of annotated pediatric data. Hence, we propose to train a segmentation model on multiple datasets, arising from different parts of the anatomy, in a multi-task and multi-domain learning framework. This approach allows to overcome the inherent scarcity of pediatric data while benefiting from a more robust shared representation. The proposed segmentation network comprises shared convolutional filters, domain-specific batch normalization parameters that compute the respective dataset statistics and a domain-specific segmentation layer. Furthermore, a supervised contrastive regularization is integrated to further improve generalization capabilities, by promoting intra-domain similarity and impose inter-domain margins in embedded space. We evaluate our contributions on two pediatric imaging datasets of the ankle and shoulder joints for bone segmentation. Results demonstrate that the proposed model outperforms state-of-the-art approaches.

show abstract

“…Fourure et al [29] presented an approach that is enhanced by multiple datasets to improve the semantic segmentation accuracy on the KITTI [36] dataset used for autonomous driving. To take advantage of training data from multiple datasets with different tasks including different label sets, they proposed a new selective loss function that can be integrated into deep networks.…”

Section: Introductionmentioning

confidence: 99%

A Brief Survey and an Application of Semantic Image Segmentation for Autonomous Driving

Kaymak

Uçar

2019

Smart Innovation, Systems and Technologies

View full text Add to dashboard Cite

Deep learning is a fast-growing machine learning approach to perceive and understand large amounts of data. In this paper, general information about the deep learning approach which is attracted much attention in the field of machine learning is given in recent years and an application about semantic image segmentation is carried out in order to help autonomous driving of autonomous vehicles. This application is implemented with Fully Convolutional Network (FCN) architectures obtained by modifying the Convolutional Neural Network (CNN) architectures based on deep learning. Experimental studies for the application are utilized 4 different FCN architectures named FCN-AlexNet, FCN-8s, FCN-16s and FCN-32s. For the experimental studies, FCNs are first trained separately and validation accuracies of these trained network models on the used dataset is compared. In addition, image segmentation inferences are visualized to take account of how precisely FCN architectures can segment objects.

show abstract

Semantic Segmentation via Multi-task, Multi-domain Learning

Cited by 6 publications

References 16 publications

Multi-task, multi-domain learning: Application to semantic segmentation and pose regression

Multi-task, multi-domain learning: Application to semantic segmentation and pose regression

Multi-Task, Multi-Domain Deep Segmentation with Shared Representations and Contrastive Regularization for Sparse Pediatric Datasets

A Brief Survey and an Application of Semantic Image Segmentation for Autonomous Driving

Contact Info

Product

Resources

About