Invariant Information Distillation for Unsupervised Image Segmentation and Clustering

By Xu Ji, João F. Henriques, Andrea Vedaldi. 2018

69 citations recorded by the source corpus.

Subfields: Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Representation & Self-Supervised Learning

Citations in the corpus

On Mutual Information Maximization for Representation Learning (2019, background): “Other approaches, such as those presented by Hu et al. (2017) and Ji et al. (2018), can be similarly subsumed under the same objective.”
Learning deep representations by mutual information estimation and maximization (2018, methodology): “Unsupervised clustering and segmentation is attainable by maximizing the MI between images associated by transforms or spatial proximity (Ji et al., 2018).”
Self-labelling via simultaneous clustering and representation learning (2019, methodology): “In our formulation, this is easily achieved by considering multiple heads (Ji et al., 2018) h1, .”
Unsupervised Learning of Object Keypoints for Perception and Control (2019, methodology): “Both instance and semantic segmentation has been approached using supervised [22, 25] and unsupervised learning [2, 9, 14, 10, 16, 21, 6] methods.”
Stacked Capsule Autoencoders (2019, methodology): “…20.8 12.5 AE (Bengio et al., 2007) § 81.2 31.4 - GAN (Radford et al., 2016) § 82.8 31.5 - IMSAT (Hu et al., 2017) † , ∇ 98.4 (0.4) 45.6 (0.8) 57.3 (3.9) IIC (Ji et al., 2018) § , † we train SCAE on MNIST , SVHN 8 and CIFAR 10 and try to assign class labels to vectors of object capsule presences.”
Using Self-Supervised Learning Can Improve Model Robustness and Uncertainty (2019, methodology): “A recent self-supervised technique is Invariant Information Clustering (IIC) [Ji et al., 2018] which teaches networks to cluster images without labels but instead by learning representations which are invariant to geometric perturbations such as rotations, scaling, and skewing.”
Audio-Visual Instance Discrimination with Cross-Modal Agreement (2020, methodology): “Self-supervised methods often try to reconstruct the input data or impose constraints on the representation, such as sparsity [48, 59, 60], noise [82] or invariance [8, 10, 11, 17, 28, 35, 53, 69] to learn a useful and transferable feature representation.”
Deep Semi-Supervised Anomaly Detection (2019, methodology): “Examples where the Infomax principle has been applied include tasks such as independent component analysis (Bell & Sejnowski, 1995), clustering (Slonim et al., 2005; Ji et al., 2018), generative modeling (Chen et al.”
Unsupervised Object Segmentation by Redrawing (2019, methodology): “2019 [26] whose method finds clusters of pixels using a learned distance invariant to some known properties.”
Semi-supervised semantic segmentation needs strong, varied perturbations (2019, methodology): “er-pixel consistency loss so that it is only applied to pixels that are not transformed outside the bounds of the image by the afﬁne transformations used in either path. A similar approach is used in [13]. We used a consistency loss weight of 0.3. 4.2.2 CutMix for segmentation The CutMix algorithm w”
Self-Supervised Learning Aided Class-Incremental Lifelong Learning (2020, background): “Recent literatures [13, 15, 17, 32, 48], have proven that semantic representations can be learned by predicting labels obtained from the input signals without any human annotations.”
Subclass Distillation (2020, methodology): “We also measure the accuracy of For comparison, the state-of-the-art (Ji et al., 2018) on fully unsupervised classiﬁcation on CIFAR-10 is 57.6% using the invariant information clustering method (IIC).”