Parallel Three-Layer Coupled Bi-GRU Algorithm for Image Caption Generation

To address the limitation of Bi-GRU models in capturing fine-grained semantic information when relying solely on forward and backward text modeling, this paper proposes a three-layer parallel structure to capture semantic information more effectively. Each of the three GRU layers extracts semantic information at different scales, enriching the semantic depth of the generated image captions. Furthermore, to integrate the semantic information captured by the three parallel GRU layers, a coupling mechanism is designed to merge the parallel hidden states, allowing interlayer interactions at each time step to enhance the semantic granularity of the fused image captions. Finally, extensive experiments on public datasets are conducted to verify the model’s effectiveness, and the experimental results are analyzed to evaluate the model’s feasibility comprehensively.