Deep Global-sense Hard-negative Discriminative Generation Hashing for Cross-modal Retrieval

Hard negative generation (HNG) provides valuable signals for deep learning, but existing methods mostly rely on local correlations while neglecting the global geometry of the embedding space. This limitation often leads to weak discrimination, particularly in cross-modal hashing, which learns compact binary codes. We propose Deep Global-sense Hard-negative Discriminative Generation Hashing (DGHDGH), a framework that constructs a structured graph with dualiterative message propagation to capture global correlations, and then performs difficulty-adaptive, channel-wise interpolation to synthesize semantically consistent hard negatives aligned with global Hamming geometry. Our approach yields more informative negatives, sharpens semantic boundaries in the Hamming cospace, and substantially enhances cross-modal retrieval. Experiments on multiple benchmarks consistently demonstrate improvements in retrieval accuracy, verifying the discriminative advantages brought by global-sense HNG in crossmodal hashing. Related code and data are available at https://github. com/QinLab-WFU/DGHDGH . Figure 1: Traditional generation methods only interpolate based on the correlation between single anchor-negative pairs, which damages the global distribution relationship of heterogeneous samples in the embedding co-space. Through the interpolation of hard negative samples with global awareness of sample correlation, the generated samples are controlled to avoid violating the feature distribution in the embedding space, which makes the co-space more discriminative.