Data Distribution Distilled Generative Model for Generalized Zero-Shot Recognition
This work addresses a bias problem in zero-shot learning for AI researchers, but it is incremental as it builds on existing generative frameworks.
The paper tackles biases in Generalized Zero-Shot Learning (GZSL) models that favor seen data by introducing D$^3$GZSL, an end-to-end generative framework that treats seen and synthesized unseen data as in-distribution and out-of-distribution, respectively, to achieve a more balanced model, and it elevates the performance of existing generative GZSL methods across established benchmarks.
In the realm of Zero-Shot Learning (ZSL), we address biases in Generalized Zero-Shot Learning (GZSL) models, which favor seen data. To counter this, we introduce an end-to-end generative GZSL framework called D$^3$GZSL. This framework respects seen and synthesized unseen data as in-distribution and out-of-distribution data, respectively, for a more balanced model. D$^3$GZSL comprises two core modules: in-distribution dual space distillation (ID$^2$SD) and out-of-distribution batch distillation (O$^2$DBD). ID$^2$SD aligns teacher-student outcomes in embedding and label spaces, enhancing learning coherence. O$^2$DBD introduces low-dimensional out-of-distribution representations per batch sample, capturing shared structures between seen and unseen categories. Our approach demonstrates its effectiveness across established GZSL benchmarks, seamlessly integrating into mainstream generative frameworks. Extensive experiments consistently showcase that D$^3$GZSL elevates the performance of existing generative GZSL methods, underscoring its potential to refine zero-shot learning practices.The code is available at: https://github.com/PJBQ/D3GZSL.git