PiperNet: a hybrid deep learning approach for monitoring papaya seed adulteration in black pepper using hyperspectral...

Adulteration of spices is a growing concern, compromising food quality, consumer safety, and industry trust. Black pepper (Piper nigrum), a high-value spice, is frequently adulterated with dried papaya seeds, which closely resemble pepper but lack nutritional value and may cause digestive issues. Conventional inspection methods often fail to detect such visually similar adulterants, highlighting the need for advanced alternatives. This study proposes a hyperspectral imaging (HSI)-based deep learning framework for detecting papaya seed adulteration in black pepper. Hyperspectral images were acquired using a Resonon Pika L camera (400-1000 nm), resulting in a dataset of 600 samples across three categories: pure pepper, pure papaya seeds, and adulterated mixtures. Pre-processing involved normalisation and conversion into 100 × 100 × 300 spectral-spatial patches, followed by one-hot encoding for classification. A 2D Convolutional Neural Network (CNN) integrated with Squeeze-and-Excitation (SE) module and Bidirectional Gated Recurrent Unit (Bi-GRU) layers was designed to jointly capture spatial textures, emphasise informative spectral channels, and model sequential wavelength dependencies. The proposed model achieved 97.5% classification accuracy, outperforming conventional imaging approaches. These findings demonstrate the potential of HSI combined with deep learning for non-destructive, automated spice authentication, supporting food quality assurance, regulatory compliance, and consumer protection.