Abstract:With the popularity of photo-sharing websites, the number of web images has exploded into unseen magnitude. Annotating such large-scale data will cost huge amount of human resources and is thus unaffordable. Motivated by this challenging problem, we propose a novel sparse graph based multilabel propagation (SGMP) scheme for super large scale datasets. Both the efficacy and accuracy of the image annotation are further investigated under different graph construction strategies, where Gaussian noise and non-Gaussian sparse noise are simultaneously considered in the formulations of these strategies. Our proposed approach outperforms the state-of-the-art algorithms by focusing on: (1) For large-scale graph construction, a simple yet efficient LSH ( Locality Sensitive Hashing )-based sparse graph construction scheme is proposed to speed up the construction. We perform the multilabel propagation on this hashing-based graph construction, which is derived with LSH approach followed by sparse graph construction within the individual hashing buckets; (2) To further improve the accuracy, we propose a novel sparsity induced scalable graph construction scheme, which is based on a general sparse optimization framework. Sparsity essentially implies a very strong prior: for large scale optimization, the values of most variables shall be zeros when the solution reaches the optimum. By utilizing this prior, the solutions of large-scale sparse optimization problems can be derived by solving a series of much smaller scale subproblems; (3) For multilabel propagation, different from the traditional algorithms that propagate over individual label independently, our proposed propagation first encodes the label information of an image as a unit label confidence vector and naturally imposes inter-label constraints and manipulates labels interactively. Then, the entire propagation problem is formulated on the concept of Kullback-Leibler divergence defined on probabilistic distributions, which guides the propagation of the supervision information. Extensive experiments on the benchmark dataset NUS-WIDE with 270k images and its lite version NUS-WIDE-LITE with 56k images well demonstrate the effectiveness and scalability of the proposed multi-label propagation scheme.

Weakly Supervised Graph Propagation Towards Collective Image Parsing

Weakly Supervised Image Parsing by Discriminatively Semantic Graph Propagation

Weakly-Supervised Image Parsing Via Constructing Semantic Graphs and Hypergraphs

Weakly-supervised Scene Parsing with Multiple Contextual Cues

Efficient large-scale image annotation by probabilistic collaborative multi-label propagation.

Large-Scale Multilabel Propagation Based on Efficient Sparse Graph Construction

Efficient Region-Aware Large Graph Construction Towards Scalable Multi-Label Propagation

Weak Supervision Learning for Object Co-Segmentation

Label localization with weakly spatial constrained graph propagation.

Interaction via Bi-directional Graph of Semantic Region Affinity for Scene Parsing

Automatic non-parametric image parsing via hierarchical semantic voting based on sparse-dense reconstruction and spatial-contextual cues.

Group-Wise Semantic Mining for Weakly Supervised Semantic Segmentation

Group-Wise Learning for Weakly Supervised Semantic Segmentation

GraphNet: Learning Image Pseudo Annotations for Weakly-Supervised Semantic Segmentation.

Image Parsing by Loopy Dynamic Programming

Deep Graph Cut Network for Weakly-Supervised Semantic Segmentation

Efficient semi-supervised annotation with Proxy-based Local Consistency Propagation

Image Annotation by K NN-sparse Graph-Based Label Propagation over Noisily Tagged Web Images

Human Parsing by Weak Structural Label

Semantic Object Parsing With Graph Lstm

Weakly Supervised Semantic Segmentation via Progressive Patch Learning