Search papers, labs, and topics across Lattice.
This study introduces the Batch-Invariant Spectral Network (BISN), an innovative framework designed to enhance the robustness of near-infrared spectroscopy for insect species authentication by mitigating batch-specific spectral variations. By employing a learnable preprocessing module initialized with Savitzky-Golay filtering and an entropy-regularized adversarial objective, BISN achieves a remarkable mean leave-one-batch-out accuracy of 0.93 across three insect species, surpassing existing methods by 4%. Additionally, the model's reliance on specific biochemical absorption regions is elucidated through explainable AI, linking its predictive capabilities to known insect biochemistry.
Achieving 93% accuracy in insect authentication despite batch variations, BISN sets a new standard for robustness in near-infrared spectroscopy.
Edible insects offer an efficient source of alternative protein, requiring less land, water and emitting less greenhouse gas than conventional livestock. However, their successful integration into the food supply chain demands reliable species authentication to control allergen exposure, prevent adulteration, and meet regulatory standards. Near-infrared spectroscopy provides a rapid analytical tool, but its performance drops when applied to production batches unseen during training due to batch-to-batch variation in spectral measurements. We introduce the Batch-Invariant Spectral Network (BISN), an end-to-end framework that combines a learnable preprocessing module, initialised with Savitzky-Golay filtering, with an entropy-regularised adversarial objective to suppress batch-specific spectral variation. In contrast to Domain-Adversarial Neural Networks, which enforce domain adaptation only after feature extraction, BISN suppress batch-effects before species-specific features are learned. Using 2,700 spectra from three species (Acheta domesticus, Hermetia illucens, and Tenebrio molitor) collected across three independent production batches, BISN achieves a mean leave-one-batch-out accuracy of 0.93 (standard deviation 0.04), outperforming the strongest baseline by four percent. Further insights gained by using explainable AI confirm that model decisions consistently rely on the lipid and protein absorption regions across all folds, connecting predictive performance to known insect biochemistry. BISN addresses both cross-batch robustness and biochemical interpretability for automated insect species authentication under realistic industrial conditions. The source code and dataset are publicly available at https://github.com/majharB/bisn.
