Learning Random-Walk Label Propagation for Weakly-Supervised Semantic Segmentation

By Paul Vernaza, Manmohan Chandraker. 2017

279 citations recorded by the source corpus.

Large-scale training for semantic segmentation is challenging due to the expense of obtaining training data for this task relative to other vision tasks. We propose a novel training approach to address this difficulty. Given cheaply-obtained sparse image labelings, we propagate the sparse labels to produce guessed dense labelings. A standard CNN-based segmentation network is trained to mimic these labelings. The label-propagation process is defined via random-walk hitting probabilities, which leads to a differentiable parameterization with uncertainty estimates that are incorporated into our loss. We show that by learning the label-propagator jointly with the segmentation predictor, we are able to effectively learn semantic edges given no direct edge supervision. Experiments also show that training a segmentation network in this way outperforms the naive approach.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

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

Incremental Learning in Semantic Segmentation from Image Labels (2021, background): “To reduce the annotation cost, different types of weak supervision have been proposed: bounding boxes [18, 32], scribbles [39, 62], points [16], and image-level labels [34, 50, 52].”
Learning Pixel-Level Semantic Affinity with Image-Level Supervision for Weakly Supervised Semantic Segmentation (2018, methodology): “Successful examples of weak supervision for semantic segmentation include bounding box [6, 12, 28], scribble [18, 36], point [1], and so on.”
Weakly Supervised Learning of Instance Segmentation With Inter-Pixel Relations (2019, methodology): “Weakly Supervised Semantic Segmentation: For weak supervision of semantic segmentation, various types of weak labels such as bounding boxes [8, 40], scribbles [29, 47], and points [3] have been utilized.”
Learning Integral Objects With Intra-Class Discriminator for Weakly-Supervised Semantic Segmentation (2020, background): “Weakly-supervised semantic segmentation (WSSS) aims at learning semantic segmentation with only coarse labels, e.g., bounding boxes [7, 34], scribbles [24, 36], and image labels [38, 20, 15, 37, 40, 14].”
Self-Supervised Equivariant Attention Mechanism for Weakly Supervised Semantic Segmentation (2020, background): “Compared to fully supervised learning, WSSS uses weak labels to guide network training, e.g ., bounding boxes [7, 18], scribbles [22, 30] and image-level classiﬁcation labels [19, 25, 27].”
Causal Intervention for Weakly-Supervised Semantic Segmentation (2020, methodology): “To this end, existing methods mainly focus on generating better seed areas [25, 57, 59, 58] and expanding these seed areas [1, 2, 20, 23, 55].”
On Regularized Losses for Weakly-supervised CNN Segmentation (2018, methodology): “The main stream of weakly-supervised methods generate segmentation proposals and train with such ”fake” ground truth [17,25,26,13,14,27].”
Non-Salient Region Object Mining for Weakly Supervised Semantic Segmentation (2021, background): “Compared to bounding boxes [14, 26, 43], points [4], and scribbles [33, 46], image-level labels [29, 51, 21, 8, 23, 2, 52, 24] are the most widely used weak annotations due to their easy availability.”