On Regularized Losses for Weakly-supervised CNN Segmentation

By Meng Tang, Federico Perazzi, Abdelaziz Djelouah, Ismail Ben Ayed, Christopher Schroers, Yuri Boykov. European Conference on Computer Vision, 2018

327 citations recorded by the source corpus.

Minimization of regularized losses is a principled approach to weak supervision well-established in deep learning, in general. However, it is largely overlooked in semantic segmentation currently dominated by methods mimicking full supervision via "fake" fully-labeled training masks (proposals) generated from available partial input. To obtain such full masks the typical methods explicitly use standard regularization techniques for "shallow" segmentation, e.g. graph cuts or dense CRFs. In contrast, we integrate such standard regularizers directly into the loss functions over partial input. This approach simplifies weakly-supervised training by avoiding extra MRF/CRF inference steps or layers explicitly generating full masks, while improving both the quality and efficiency of training. This paper proposes and experimentally compares different losses integrating MRF/CRF regularization terms. We juxtapose our regularized losses with earlier proposal-generation methods using explicit regularization steps or layers. Our approach achieves state-of-the-art accuracy in semantic segmentation with near full-supervision quality.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/36ad8eaa6d01ceec2efda3a9563320822920088f

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

Single-Stage Semantic Segmentation From Image Labels (2020, background): “This problem setup is especially challenging compared to other weakly supervised scenarios that assume available localisation cues, such as bounding boxes, scribbles, and points [4, 10, 27, 33, 51].”
FickleNet: Weakly and Semi-Supervised Semantic Image Segmentation Using Stochastic Inference (2019, methodology): “The methods of recent introduction have achieved successful results using annotations that provide location information such as scribbles or bounding boxes [4, 29].”
Semi-Supervised Semantic Segmentation With High- and Low-Level Consistency (2019, methodology): “Attempting to alleviate the problem, several methods exploit weaker forms of supervision: image-level labels [1], [32], [38], bounding boxes [7], [29], or scribbles [20], [36].”
Embracing Imperfect Datasets: A Review of Deep Learning Solutions for Medical Image Segmentation (2019, methodology): “Tang et al. (2018) train a stacked GAN (SGAN) to pre-process CT images, where the first GAN generates a denoised image and the second GAN generates a high resolution image.”
Efficient Segmentation: Learning Downsampling Near Semantic Boundaries (2019, methodology): “Alternatively, one can directly use objective (2) as a regularized loss function [51, 52].”
Multi-class Token Transformer for Weakly Supervised Semantic Segmentation (2022, background): “To tackle this problem, speciﬁc segmentation losses have been proposed to cater for deﬁcient segmentation supervision, including SEC loss [19], CRF loss [35, 47] and contrastive loss [17].”
DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision (2021, unknown): “A number of methods have been proposed to learn semantic segmentation with image-level class labels [69–73], points [74,75], scribbles [76–79] and bounding boxes [46,80–83].”