Deep Co-Training for Semi-Supervised Image Segmentation

Peng, Jizong; Estrada, Guillermo; Pedersoli, Marco; Desrosiers, Christian

doi:10.48550/arxiv.1903.11233

Cited by 5 publications

(9 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Semi-supervised segmentation: The bulk of semisupervised methods for segmentation can be roughly grouped into four categories: self-training methods [9]- [11], regularization methods [12]- [18], adversarial learning [19]- [22], and cotraining methods [5], [6], [8]. In basic self-training approaches, a model generates pseudo-labels for unlabeled data and is then retained with the updated set of labeled examples.…”

Section: Related Workmentioning

confidence: 99%

“…where adversarial examples built from training images are used to enforce diversity among different classifiers. This idea was later extended to medical image segmentation by Peng et al in [8]. More recently, Xia et al [6] introduced an uncertaintyaware multi-view co-training framework in which prediction uncertainty estimated using a Bayesian approach is employed to merge the output for different views.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Self-paced and self-consistent co-training for semi-supervised image segmentation

Wang

Peng

Pedersoli

et al. 2021

Medical Image Analysis

Self Cite

View full text Add to dashboard Cite

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Self-paced and self-consistent co-training for semi-supervised image segmentation

Wang

Peng

Pedersoli

et al. 2021

Medical Image Analysis

Self Cite

View full text Add to dashboard Cite

“…Semi-supervised segmentation Although initially developed for classification (Oliver et al, 2018), a wide range of semi-supervised methods have also been proposed for semantic segmentation. These methods are based on various learning techniques, including selftraining (Bai et al, 2017), distillation (Radosavovic et al, 2018), attention learning (Min and Chen, 2018), adversarial learning (Souly et al, 2017;Zhang et al, 2017), entropy minimization (Vu et al, 2019), co-training (Peng et al, 2019b;Zhou et al, 2019), temporal ensembling (Perone and Cohen-Adad, 2018), manifold learning (Baur et al, 2017), and data augmentation (Chaitanya et al, 2019;Zhao et al, 2019a). Among recently proposed methods, consistency-based regularization has emerged as an effective way to improve performance by enforcing the network to output similar predictions for unlabeled images under different transformations (Bortsova et al, 2019).…”

Section: Related Workmentioning

confidence: 99%

Boosting Semi-supervised Image Segmentation with Global and Local Mutual Information Regularization

Peng,

Pedersoli,

Desrosiers

2021

Preprint

Self Cite

View full text Add to dashboard Cite

The scarcity of labeled data often impedes the application of deep learning to the segmentation of medical images. Semi-supervised learning seeks to overcome this limitation by leveraging unlabeled examples in the learning process. In this paper, we present a novel semi-supervised segmentation method that leverages mutual information (MI) on categorical distributions to achieve both global representation invariance and local smoothness. In this method, we maximize the MI for intermediate feature embeddings that are taken from both the encoder and decoder of a segmentation network. We first propose a global MI loss constraining the encoder to learn an image representation that is invariant to geometric transformations. Instead of resorting to computationally-expensive techniques for estimating the MI on continuous feature embeddings, we use projection heads to map them to a discrete cluster assignment where MI can be computed efficiently. Our method also includes a local MI loss to promote spatial consistency in the feature maps of the decoder and provide a smoother segmentation. Since mutual information does not require a strict ordering of clusters in two different assignments, we incorporate a final consistency regularization loss on the output which helps align the cluster labels throughout the network. We evaluate the method on three challenging publicly-available datasets for medical image segmentation. Experimental results show our method to outperform recently-proposed approaches for semi-supervised segmentation and provide an accuracy near to full supervision while training with very few annotated images

show abstract

“…Given that two independent views were not available, the authors instead generated adversarial examples to enforce diversity of the two classifiers. Peng et al [13] extended the previous work in order to carry out semantic segmentation of medical images. The co-training algorithm was also used in [14] for the task of object classification in RGB-D images acquired with a Kinect sensor.…”

Section: Related Workmentioning

confidence: 99%

Lidar-Camera Co-Training for Semi-Supervised Road Detection

Caltagirone,

Svensson,

Wahde

et al. 2019

Preprint

View full text Add to dashboard Cite

Recent advances in the field of machine learning and computer vision have enabled the development of fast and accurate road detectors. Commonly such systems are trained within a supervised learning paradigm where both an input sensor's data and the corresponding ground truth label must be provided. The task of generating labels is commonly carried out by human annotators and it is notoriously time consuming and expensive. In this work, it is shown that a semi-supervised approach known as co-training can provide significant F1-score average improvements compared to supervised learning. In co-training, two classifiers acting on different views of the data cooperatively improve each other's performance by leveraging unlabeled examples. Depending on the amount of labeled data used, the improvements ranged from 1.12 to 6.10 percentage points for a camera-based road detector and from 1.04 to 8.14 percentage points for a lidar-based road detector. Lastly, the co-training algorithm is validated on the KITTI road benchmark, achieving high performance using only 36 labeled training examples together with several thousands unlabeled ones.

show abstract

Deep Co-Training for Semi-Supervised Image Segmentation

Cited by 5 publications

References 38 publications

Self-paced and self-consistent co-training for semi-supervised image segmentation

Self-paced and self-consistent co-training for semi-supervised image segmentation

Boosting Semi-supervised Image Segmentation with Global and Local Mutual Information Regularization

Lidar-Camera Co-Training for Semi-Supervised Road Detection

Contact Info

Product

Resources

About