Guided Collaborative Training for Pixel-wise Semi-Supervised Learning

By Zhanghan Ke, Kaican Li, Qiong Yan, Rynson W. H. Lau. European Conference on Computer Vision, 2020

296 citations recorded by the source corpus.

We investigate the generalization of semi-supervised learning (SSL) to diverse pixel-wise tasks. Although SSL methods have achieved impressive results in image classification, the performances of applying them to pixel-wise tasks are unsatisfactory due to their need for dense outputs. In addition, existing pixel-wise SSL approaches are only suitable for certain tasks as they usually require to use task-specific properties. In this paper, we present a new SSL framework, named Guided Collaborative Training (GCT), for pixel-wise tasks, with two main technical contributions. First, GCT addresses the issues caused by the dense outputs through a novel flaw detector. Second, the modules in GCT learn from unlabeled data collaboratively through two newly proposed constraints that are independent of task-specific properties. As a result, GCT can be applied to a wide range of pixel-wise tasks without structural adaptation. Our extensive experiments on four challenging vision tasks, including semantic segmentation, real image denoising, portrait image matting, and night image enhancement, show that GCT outperforms state-of-the-art SSL methods by a large margin. Our code available at: this https URL.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/e8271568a813ee7ce25d66c6a8cb04e71e590b75

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Semi-Supervised Semantic Segmentation with Cross Pseudo Supervision (2021, methodology): “We follow the partition protocols of Guided Collaborative Training (GCT) [17] and divide the whole training set to two groups via randomly sub-sampling 1/2, 1/4, 1/8 and 1/16 of the whole set as the labeled set and regard the remaining images as the unlabeled set.”
DatasetGAN: Efficient Labeled Data Factory with Minimal Human Effort (2021, background): “Semi-Supervised Learning: Given a large set of unlabeled images and a small set of annotated images, semisupervised approaches [37, 47, 22, 40, 27] aim to learn better segmentation networks than with supervised data alone.”
ST++: Make Self-trainingWork Better for Semi-supervised Semantic Segmentation (2021, methodology): “The DeepLabv2 model is initialized with the MS COCO [36] pre-trained parameters following [28], while the other backbones are pre-trained on ImageNet [14].”
Revisiting Weak-to-Strong Consistency in Semi-Supervised Semantic Segmentation (2022, methodology): “Motivated by the rapid progress in SSL, recent methods [49, 18, 32, 46, 81, 79, 78, 1, 75, 40, 42, 34, 74] strive for simpler training paradigms from the perspective of consistency regularization and entropy minimization.”
Semantic Segmentation with Generative Models: Semi-Supervised Learning and Strong Out-of-Domain Generalization (2021, methodology): “We also benchmark our approach against several state-of-the-art SLL methods for segmentation that have code available: the mean teacher model with transformation-consistency (MT) [87, 53], the adversarial training-based method [37] (AdvSSL), and also the recently proposed Guided Collaborative Traini”
Semi-Supervised and Unsupervised Deep Visual Learning: A Survey (2022, methodology): “In a similar spirit as the standard consistency regularization in SSL, recent works in semi-supervised semantic segmentation [31], [178], [179], [180] resort to enforcing pixel consistency among the images before and after perturbations, whilst perturbations being introduced at the input space…”
Bootstrapping Semantic Segmentation with Regional Contrast (2021, background): “Some solutions improved the quality of pseudo-labelling, using adversarial learning (Hung et al., 2019; Mittal et al., 2019) or enforcing consistency from different augmented images (French et al., 2020; Olsson et al., 2021).”
PseudoSeg: Designing Pseudo Labels for Semantic Segmentation (2020, background): “Consistency regularization based approaches have also been proposed recently, by enforcing the predictions to be consistent, either from augmented input images (French et al., 2020; Kim et al., 2020), perturbed feature embeddings (Ouali et al., 2020), or different networks (Ke et al., 2020).”