One-Shot Learning for Semantic Segmentation
By Amirreza Shaban, Shray Bansal, Z. Liu, Irfan Essa, Byron Boots. 2017
897 citations recorded by the source corpus.
Low-shot learning methods for image classification support learning from sparse data. We extend these techniques to support dense semantic image segmentation. Specifically, we train a network that, given a small set of annotated images, produces parameters for a Fully Convolutional Network (FCN). We use this FCN to perform dense pixel-level prediction on a test image for the new semantic class. Our architecture shows a 25% relative meanIoU improvement compared to the best baseline methods for one-shot segmentation on unseen classes in the PASCAL VOC 2012 dataset and is at least 3 times faster.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging, Meta & Continual Learning
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
- Trevor Darrell
- Bernt Schiele
- Sergey Levine
- Jiashi Feng
- Dacheng Tao
- Yi Yang
- Yoshua Bengio
- Chunhua Shen
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Ming-Hsuan Yang
- Trevor Darrell
- Yi Yang
- Bernt Schiele
- Xiaogang Wang
- Chunhua Shen
In Radiology, Nuclear Medicine and Imaging
In Meta & Continual Learning
Citations in the corpus
- Language-driven Semantic Segmentation (2022, methodology): “Few-shot approaches (Shaban et al., 2017; Rakelly et al., 2018; Siam et al., 2019; Wang et al., 2019; Zhang et al., 2019; Nguyen & Todorovic, 2019; Liu et al., 2020b; Wang et al., 2020; Tian et al., 2020; Boudiaf et al., 2021; Min et al., 2021) offer ways to learn to segment novel classes based on…”
- Incremental Few-Shot Semantic Segmentation via Embedding Adaptive-Update and Hyper-class Representation (2022, methodology): “PASCAL-5 .”
- Meta-Learning in Neural Networks: A Survey (2020, methodology): “Hypernetworkbased meta-learners have been applied in the one-shot regime [207], and performance was later improved by”
- CRNet: Cross-Reference Networks for Few-Shot Segmentation (2020, methodology): “The class division of the PASCAL VOC 2012 dataset proposed in [17].”
- AMP: Adaptive Masked Proxies for Few-Shot Segmentation (2019, methodology): “proposed a 2-branch method [28], where the second branch predicts the parameters for the final segmentation layer.”
- Few-Shot Semantic Segmentation with Prototype Learning (2018, methodology): “To evaluate the performance of our framework, we experiment on PASCAL-5 i [31], which is based on PASCAL VOC 2012 dataset [11].”
- Image Segmentation Using Text and Image Prompts (2021, methodology): “In conventional, CNN-based one-shot semantic segmentation, masked pooling [31] has emerged as a standard technique to compute a prototype vector for conditioning.”
- Learning What Not to Segment: A New Perspective on Few-Shot Segmentation (2022, methodology): “Moreover, we also make comparisons of our model with other advanced approaches in terms of FB-IoU on PASCAL-5 i (see Tab.”
- A Comprehensive Survey of Few-shot Learning: Evolution, Applications, Challenges, and Opportunities (2022, methodology): “Few-shot semantic segmentation was first proposed in [31] until 2017.”
- Semantic Projection Network for Zero- and Few-Label Semantic Segmentation (2019, methodology): “Prior models that tackle few-shot semantic segmentation [42, 11] operate in the foreground-background segmentation setting.”
- Small Sample Learning in Big Data Era (2018, background): “…future, more other problems are worthy to be investigated, such as generation (Lake et al., 2015), synthesis (Long et al., 2017c), parsing (Zhu et al., 2018a; Bansal et al., 2018; Rahman et al., 2018; Shaban et al., 2017), account for concepts are far more complicated than object categories only.”
- Hypercorrelation Squeeze for Few-Shot Segmenation (2021, methodology): “[61] propose one-shot semantic segmentation networks which (meta-) learns to generate parameters of FCN [62].”
- Feature Weighting and Boosting for Few-Shot Segmentation (2019, methodology): “For PASCAL-5, we use the same 4-fold cross-validation setup as prior work [26, 20, 6].”
- One-Shot Instance Segmentation (2018, methodology): “A range of one-shot segmentation tasks exist, including one-shot semantic segmentation [56, 43, 12, 37], texture segmentation [61], medical image segmentation [67] and recent work on cosegmentation [28]2.”
- Prior Guided Feature Enrichment Network for Few-Shot Segmentation (2020, neutral): “Feature processing can be accomplished by generating weights for the classifier [29], cosine-similarity calculation [4], [40], or convolutions [13], [46] to generate the final prediction.”