Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Food Quality Laboratory » Research » Publications at this Location » Publication #397997
Research Project: Reducing Postharvest Loss and Improving Fresh Produce Marketability and Nutritive Values through Technological Innovations and Process Optimization

Location: Food Quality Laboratory

Title: SQ-Swin: Siamese Quadratic Swin transformer for lettuce browning prediction

Author
item WANG, DAYANG - UNIVERSITY OF MASSACHUSETTS
item Luo, Yaguang - Sunny
item ZHANG, BOCE - UNIVERSITY OF MASSACHUSETTS
item XU, YONGSHUN - UNIVERSITY OF MASSACHUSETTS
item YU, HENGYONG - UNIVERSITY OF MASSACHUSETTS

Submitted to: IEEE Access
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/8/2023
Publication Date: 11/13/2023
Citation: Wang, D., Luo, Y., Zhang, B., Xu, Y., Yu, H. 2023. SQ-Swin: Siamese Quadratic Swin transformer for lettuce browning prediction. IEEE Access. 11:128724-128735. https://doi.org/10.1109/ACCESS.2023.3332488.
DOI: https://doi.org/10.1109/ACCESS.2023.3332488

Interpretive Summary: Developing browning resistant lettuce cultivars is a cost- effective approach to mitigate enzymatic browning, a major quality defect of fresh-cut lettuce. However, existing methods to assess browning are labor intense and subjective. Via a multidisciplinary collaboration, food scientists, engineers, and AI specialists in USDA-ARS, University of Florida, and University of Massachusetts Lowell investigated the possibility of using computer vision and deep learning to predict browning among lettuce cultivars. Using a pretrained Siamese Quadratic Swin transformer (SQ-Swin), the team demonstrated that this SQ-Swin program outperformed the traditional methods and other deep learning-based backbones. Findings will benefit vegetable growers and processors in identifying browning resistant lettuce cultivars and other researchers in advancing artificial intelligence programs for agricultural applications.

Technical Abstract: Enzymatic browning is a major quality defect of packaged “ready-to-eat” fresh-cut lettuce salads. While there have been many research and breeding efforts to counter this problem, progress is hindered by the lack of a technology to rapidly, objectively, and reliably identify and quantify browning. Here, we report a deep learning model for lettuce browning prediction. To the best of our knowledge, it is the first-of-its-kind on deep learning for lettuce browning prediction using a pretrained Siamese Quadratic Swin (SQ-Swin) transformer with several highlights. First, our model includes quadratic features in the transformer model which is more powerful to incorporate real-world representations than the linear transformer. Second, a multi-scale training strategy is proposed to augment the data and explore more of the inherent self-similarity of the lettuce images. Third, the proposed model uses a Siamese architecture which learns the inter-relations among the limited training samples. Fourth, the model is pretrained on the ImageNet and then trained with the reptile meta-learning algorithm to learn higher-order gradients than a regular one. Experiment results on the fresh-cut lettuce datasets show that the proposed SQ-Swin outperforms the traditional methods and other deep learning-based backbones.