Deployment and analysis of instance segmentation algorithm for in-field yield estimation of sweet potatoes

Shape estimation of sweetpotato (SP) storage roots is inherently challenging due to their varied size and shape characteristics. Even measuring “simple” metrics, such as length and diameter, requires significant time investments either directly in-field or afterward using automated graders. We present the results of a model that can perform grading and provide yield estimates directly in the field faster than manual measurements. Detectron2, a library consisting of deep-learning object detection algorithms, was used to implement Mask R-CNN, an instance segmentation model.  This model was deployed for in-field grade estimation of SP roots and evaluated against an optical sorter.  Roots from various clones imaged with a cellphone during trials between 2019 and 2020, were used in the model’s training and validation to fine-tune a model to detect SP roots.  Our results showed that the model (Average Precision = 74.1) could distinguish individual roots in environmental conditions, including variations in lighting and soil characteristics.  Root mean square error (RMSE) for length, diameter, and weight, from the model compared to a commercial optical sorter, were 0.66 cm, 1.22 cm, and 74.73 g, respectively, while the RMSE of root counts per plot was 5.27 roots, with R^2 = 0.8. This phenotyping strategy has the potential to enable rapid yield estimates in the field without the need for sophisticated and costly sorters and may be more readily deployed in environments with limited access to these resources or facilities.