Deployment and analysis of instance segmentation algorithm for in‐field yield estimation of sweet potatoes
Shape estimation of sweet potato (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 pres...
Uložené v:
| Vydané v: | Plant phenome journal Ročník 8; číslo 1 |
|---|---|
| Hlavní autori: | , , , , , , , , |
| Médium: | Journal Article |
| Jazyk: | English |
| Vydavateľské údaje: |
01.12.2025
|
| ISSN: | 2578-2703, 2578-2703 |
| On-line prístup: | Získať plný text |
| Tagy: |
Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
|
| Shrnutí: | Shape estimation of sweet potato (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 region‐based convolutional neural network, 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 cell phone 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 = 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.
Developed computer vision pipeline to measure sweetpotato roots dimensions. Simulated variation in perceived sweetpotato shape. Validated pipeline using large dataset of sweetpotato roots measured using commercial systems.
Sweetpotatoes vary greatly in shape and size, making measurements slow. This study tested if phone pictures and an AI system could quickly estimate sweetpotato count, size, and weight directly in the field. The AI was trained using photos of harvested sweetpotatoes. It successfully identified individual roots in different conditions. Compared to expensive sorting machines, the AI estimated root counts well (coefficient of determination 0.8). It estimated length with about 0.7 cm error, diameter with 1.2 cm error, and weight with 75 g error. This phone‐based AI method offers a potential low‐cost, portable tool for faster yield estimates in the field, helping farmers and breeders, especially those with limited resources |
|---|---|
| ISSN: | 2578-2703 2578-2703 |
| DOI: | 10.1002/ppj2.70026 |