EUVIMEDEUVIMED entdecken
Crossref

Primitive Memory Networks With Qualified Prompts for Compositional Zero-Shot Learning

Yutang Jin, Boyu Zhao, Shitao Chen, Shiming Chen, Guo-Sen Xie, Yisong Wang, Xinge You

IEEE Transactions on Image Processing · 2026 · 35 · 11628-11641

DOI: 10.1109/tip.2026.3748354 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Image Processing, 2026. Verfasst von Yutang Jin und 6 weiteren Autor:innen. Die Arbeit nennt 65 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Hubei Provincial Natural Science Foundation of China, Qiankehe Chengguo DXGA (2025) Key Project.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Compositional Zero-Shot Learning (CZSL) recognizes unseen attribute-object compositions by transferring primitive knowledge learned from seen pairs. Existing CLIP-based approaches often rely on simple textual prompts and limited visual representatives, which may inadequately characterize primitives with diverse semantics and intra-class appearances. To address these limitations, we propose Primitive Memory Networks with Qualified Prompts (PMNQP) to provide complementary semantic and visual evidence for CZSL. Specifically, qualified prompts incorporate multiple primitive-related descriptions generated once by a large language model offline, without requiring LLM access during model training or inference. Furthermore, primitive memory networks maintain a fixed-capacity bank of selected visual exemplars for each attribute and object, update slot membership online during training, and perform query-dependent readout from the fixed banks at inference. Together, these complementary semantic and visual pathways address the limited context provided by coarse primitive prompts and the inadequate coverage of intra-class variation by a single visual representation. Extensive experiments on three benchmarks show that PMNQP achieves competitive performance under both closed-world and open-world CZSL settings. Code is available at https://github.com/YtJin-git/PMNQP.

Förderung

  • National Natural Science Foundation of China
  • Hubei Provincial Natural Science Foundation of China
  • Qiankehe Chengguo DXGA (2025) Key Project

Zitieren

APA

Jin, Y., Zhao, B., Chen, S., Chen, S., Xie, G. S., Wang, Y., & You, X. (2026). Primitive Memory Networks With Qualified Prompts for Compositional Zero-Shot Learning. IEEE Transactions on Image Processing, 35, 11628-11641. https://doi.org/10.1109/tip.2026.3748354

Vancouver

Jin Y, Zhao B, Chen S, Chen S, Xie GS, Wang Y, et al. Primitive Memory Networks With Qualified Prompts for Compositional Zero-Shot Learning. IEEE Transactions on Image Processing. 2026;35:11628-11641. doi:10.1109/tip.2026.3748354

BibTeX
@article{Jin2026,
  author = {Jin, Yutang and Zhao, Boyu and Chen, Shitao and Chen, Shiming and Xie, Guo-Sen and Wang, Yisong and You, Xinge},
  title = {Primitive Memory Networks With Qualified Prompts for Compositional Zero-Shot Learning},
  journal = {IEEE Transactions on Image Processing},
  year = {2026},
  volume = {35},
  pages = {11628--11641},
  doi = {10.1109/tip.2026.3748354}
}
Kostenlos starten →