Hybrid Models for Open Set Recognition
By Hongjie Zhang, Ang Li, Jie Guo, Yanwen Guo. 2020
214 citations recorded by the source corpus.
Open set recognition requires a classifier to detect samples not belonging to any of the classes in its training set. Existing methods fit a probability distribution to the training samples on their embedding space and detect outliers according to this distribution. The embedding space is often obtained from a discriminative classifier. However, such discriminative representation focuses only on known classes, which may not be critical for distinguishing the unknown classes. We argue that the representation space should be jointly learned from the inlier classifier and the density estimator (served as an outlier detector). We propose the OpenHybrid framework, which is composed of an encoder to encode the input data into a joint embedding space, a classifier to classify samples to inlier classes, and a flow-based density estimator to detect whether a sample belongs to the unknown category. A typical problem of existing flow-based models is that they may assign a higher likelihood to outliers. However, we empirically observe that such an issue does not occur in our experiments when learning a joint representation for discriminative and generative components. Experiments on standard open set benchmarks also reveal that an end-to-end trained OpenHybrid model significantly outperforms state-of-the-art methods and flow-based baselines.
Subfields: Computer Vision and Pattern Recognition, Anomaly & Outlier Detection
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Andrea Vedaldi
- Andrew Zisserman
- Ming-Hsuan Yang
- Jitendra Malik
- Dacheng Tao
- Trevor Darrell
- Yi Yang
In Anomaly & Outlier Detection
Citations in the corpus
- Language-driven Semantic Segmentation (2022, background): “A variety of solutions have been proposed (Zhang et al., 2020b; Liu et al., 2020a; Perera et al., 2020; Zhou et al., 2021) for open-set recognition (Scheirer et al., 2012; Geng et al., 2020).”
- Adversarial vulnerability of powerful near out-of-distribution detection (2022, background): “Improving OOD detection has seen progress by training generative models [6; 7; 2; 8], and modifying objective and loss functions [9].”
- Adversarial Reciprocal Points Learning for Open Set Recognition (2021, background): “[21] proposed a composed of classifier and a flow-based density estimator into a joint embedding space.”
- Generalized Category Discovery (2022, background): “[47] jointly trains a flowbased density estimator and a classification based encoder”
- Contrastive Training for Improved Out-of-Distribution Detection (2020, background): “[36] show that scoring with a deep flow-based network can also significantly improve performance.”
- A Unifying Review of Deep and Shallow Anomaly Detection (2020, background): “Nonetheless, researchers have developed a separate set of methods for open set recognition [240], [497]–[501], and an important goal for future research is to evaluate these methods from the AD perspective and to evaluate AD algorithms from the open set perspective.”
- OpenGAN: Open-Set Recognition via Open Data Generation (2021, unknown): “To recognize open-set examples, they resort to post-hoc functions like density estimation [59, 57], uncertainty modeling [17, 29], and input image reconstruction [39, 20, 14, 50].”
- Exploring the Limits of Out-of-Distribution Detection (2021, methodology): “There has been a lot of progress in detecting OOD inputs including methods based on discriminative models [Hendrycks and Gimpel, 2016, Lee et al., 2018, Liang et al., 2017, Liu et al., 2020] as well as methods based on deep generative models [Nalisnick et al., 2019, Zhang et al., 2020].”
- A Simple Fix to Mahalanobis Distance for Improving Near-OOD Detection (2021, background): “A lot of progress has been made in improving OOD detection by training complicated generative models [2; 15; 20; 14], modifying objective functions [22], and exposing to OOD samples while training [8].”