A Decomposable Causal View of Compositional Zero-Shot Learning

Yang, Muli; Xu, Chenghao; Wu, Aming; Deng, Cheng

doi:10.1109/tmm.2022.3200578

Cited by 16 publications

(11 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Compositional Zero-shot Learning (CZSL). CZSL [5,9,18,25,26,43] has two main strategies for inferring unseen compositions. The first assumes that unseen and seen compositions share the same attribute and object scopes.…”

Section: Related Workmentioning

confidence: 99%

Diversity-Boosted Generalization-Specialization Balancing for Zero-Shot Learning

Liu

Chang

et al. 2023

IEEE Trans. Multimedia

View full text Add to dashboard Cite

The objective of Active Learning is to strategically label a subset of the dataset to maximize performance within a predetermined labeling budget. In this study, we harness features acquired through self-supervised learning. We introduce a straightforward yet potent metric, Cluster Distance Difference, to identify diverse data. Subsequently, we introduce a novel framework, Balancing Active Learning (BAL), which constructs adaptive sub-pools to balance diverse and uncertain data. Our approach outperforms all established active learning methods on widely recognized benchmarks by 1.20%. Moreover, we assess the efficacy of our proposed framework under extended settings, encompassing both larger and smaller labeling budgets. Experimental results demonstrate that, when labeling 80% of the samples, the performance of the current SOTA method declines by 0.74%, whereas our proposed BAL achieves performance comparable to the full dataset. Codes are available at https://github.com/JulietLJY/BAL.

show abstract

Section: Related Workmentioning

confidence: 99%

Diversity-Boosted Generalization-Specialization Balancing for Zero-Shot Learning

Liu

Chang

et al. 2023

IEEE Trans. Multimedia

View full text Add to dashboard Cite

show abstract

“…Compositional learning [25], [30], [31], [32], [33], [34], [35] aims to enable models to learn simple primitives in images. Compositional zero-shot learning further recognizes unseen compositions in images composed of seen simple primitives.…”

Section: Compositional Zero-shot Learningmentioning

confidence: 99%

“…Geng et al [50] disentangle knowledge graphs into ontology embeddings to capture finegrained semantic information. In compositional zero-shot learning, Yang et al [25] disentangle images into states and objects based on the causal effects in compositions. Saini et al [28] apply a state/object affinity network to disentangle the same states or objects from image pairs contrastively.…”

Section: Knowledge Disentanglement In Zero-shot Learningmentioning

confidence: 99%

“…In other words, states and objects lead to visual changes to each other. Simple primitives in images are entangled visually [25]. Most current works [16], [19], [25], [26], [27], [28] view states or objects as the bases.…”

Section: Introductionmentioning

confidence: 99%

“…Simple primitives in images are entangled visually [25]. Most current works [16], [19], [25], [26], [27], [28] view states or objects as the bases. They simulate the visual changes on bases caused by contextuality, inferring the possible entangled embedding.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Simple Primitives with Feasibility- and Contextuality-Dependence for Open-World Compositional Zero-shot Learning

Liu¹,

Li²,

Yao³

et al. 2022

Preprint

View full text Add to dashboard Cite

The task of Compositional Zero-Shot Learning (CZSL) is to recognize images of novel state-object compositions that are absent during the training stage. Previous methods of learning compositional embedding have shown effectiveness in closed-world CZSL. However, in Open-World CZSL (OW-CZSL), their performance tends to degrade significantly due to the large cardinality of possible compositions. Some recent works separately predict simple primitives (i.e., states and objects) to reduce cardinality. However, they consider simple primitives as independent probability distributions, ignoring the heavy dependence between states, objects, and compositions. In this paper, we model the dependence of compositions via feasibility and contextuality. Feasibility-dependence refers to the unequal feasibility relations between simple primitives, e.g., hairy is more feasible with cat than with building in the real world. Contextuality-dependence represents the contextual variance in images, e.g., cat shows diverse appearances under the state of dry and wet. We design Semantic Attention (SA) and generative Knowledge Disentanglement (KD) to learn the dependence of feasibility and contextuality, respectively. SA captures semantics in compositions to alleviate impossible predictions, driven by the visual similarity between simple primitives. KD disentangles images into unbiased feature representations, easing contextual bias in predictions. Moreover, we complement the current compositional probability model with feasibility and contextuality in a compatible format. Finally, we conduct comprehensive experiments to analyze and validate the superior or competitive performance of our model, Semantic Attention and knowledge Disentanglement guided Simple Primitives (SAD-SP), on three widely-used benchmark OW-CZSL datasets.

show abstract