Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Troika: Multi-Path Cross-Modal Traction for Compositional Zero-Shot Learning

About

Recent compositional zero-shot learning (CZSL) methods adapt pre-trained vision-language models (VLMs) by constructing trainable prompts only for composed state-object pairs. Relying on learning the joint representation of seen compositions, these methods ignore the explicit modeling of the state and object, thus limiting the exploitation of pre-trained knowledge and generalization to unseen compositions. With a particular focus on the universality of the solution, in this work, we propose a novel paradigm for CZSL models that establishes three identification branches (i.e., Multi-Path) to jointly model the state, object, and composition. The presented Troika is our implementation that aligns the branch-specific prompt representations with decomposed visual features. To calibrate the bias between semantically similar multi-modal representations, we further devise a Cross-Modal Traction module into Troika that shifts the prompt representation towards the current visual content. We conduct extensive experiments on three popular benchmarks, where our method significantly outperforms existing methods in both closed-world and open-world settings. The code will be available at https://github.com/bighuang624/Troika.

Siteng Huang, Biao Gong, Yutong Feng, Min Zhang, Yiliang Lv, Donglin Wang• 2023

Related benchmarks

TaskDatasetResultRank
Generalized Compositional Zero-Shot LearningC-GQA (test)
AUC0.092
46
Compositional Zero-Shot LearningUT-Zappos Closed World
HM54.6
42
Compositional Zero-Shot LearningC-GQA Closed World
HM29.4
41
Compositional Zero-Shot LearningUT-Zappos open world
HM47.8
38
Compositional Zero-Shot LearningMIT-States open world
HM20.1
38
Compositional Zero-Shot LearningC-GQA open world
HM Score10.9
35
Compositional Zero-Shot LearningMIT-States Closed World
Harmonic Mean (HM)0.393
32
Showing 7 of 7 rows

Other info

Follow for update