Published on Fri Jul 10 2020

PCAMs: Weakly Supervised Semantic Segmentation Using Point Supervision

R. Austin McEver, B. S. Manjunath

Current methods for generating semantic segmentation rely heavily on a large set of images that have each pixel labeled with a class of interest label or background. This method includes point annotations in the training of a convolutional neural network for improved localization and class activation maps.

0
0
0
Abstract

Current state of the art methods for generating semantic segmentation rely heavily on a large set of images that have each pixel labeled with a class of interest label or background. Coming up with such labels, especially in domains that require an expert to do annotations, comes at a heavy cost in time and money. Several methods have shown that we can learn semantic segmentation from less expensive image-level labels, but the effectiveness of point level labels, a healthy compromise between all pixels labelled and none, still remains largely unexplored. This paper presents a novel procedure for producing semantic segmentation from images given some point level annotations. This method includes point annotations in the training of a convolutional neural network (CNN) for producing improved localization and class activation maps. Then, we use another CNN for predicting semantic affinities in order to propagate rough class labels and create pseudo semantic segmentation labels. Finally, we propose training a CNN that is normally fully supervised using our pseudo labels in place of ground truth labels, which further improves performance and simplifies the inference process by requiring just one CNN during inference rather than two. Our method achieves state of the art results for point supervised semantic segmentation on the PASCAL VOC 2012 dataset \cite{everingham2010pascal}, even outperforming state of the art methods for stronger bounding box and squiggle supervision.

Tue Jun 16 2015
Computer Vision
Decoupled Deep Neural Network for Semi-supervised Semantic Segmentation
We propose a novel deep neural network architecture for semi-supervised semantic segmentation using heterogeneous annotations. Contrary to existing approaches, our algorithm decouples classification and segmentation, and learns a separate network for each task. Our algorithm shows outstanding performance even with much less training data.
0
0
0
Wed Mar 28 2018
Computer Vision
Learning Pixel-level Semantic Affinity with Image-level Supervision for Weakly Supervised Semantic Segmentation
Segmentation labels are one of the main obstacles to realizing semantic segmentation in the wild. We present a novel framework that generates segmentation labels of images given their image-level class labels. The semantic propagation is then realized by random walk metrics.
0
0
0
Tue Apr 13 2021
Computer Vision
All you need are a few pixels: semantic segmentation with PixelPick
0
0
0
Tue Oct 16 2018
Computer Vision
Generating Self-Guided Dense Annotations for Weakly Supervised Semantic Segmentation
Most weakly-supervised methods rely on external models to infer pseudo pixel-level labels for training semantic segmentation models. We propose a novel self-guided strategy to utilize features learned across multiple levels to progressively generate dense pseudo labels.
0
0
0
Tue Jun 11 2019
Computer Vision
Gated CRF Loss for Weakly Supervised Semantic Image Segmentation
State-of-the-art approaches for semantic segmentation rely on deep convolutional neural networks trained on fully annotated datasets. These datasets are notoriously expensive to collect, both in terms of time and money. To remedy this situation, weakly supervised methods leverage other forms of supervision that require substantially less annotation effort.
0
0
0
Mon Feb 09 2015
Computer Vision
Weakly- and Semi-Supervised Learning of a DCNN for Semantic Image Segmentation
Deep convolutional neural networks (DCNNs) trained on a large number of images with strong pixel-level annotations have recently significantly pushed the state-of-art in semantic image segmentation. We develop Expectation-Maximization (EM)methods for semantic image segments training.
0
0
0