ABSTRACT
Conventional methods of crop yield estimation are costly, inefficient, and prone to error resulting in poor yield estimates. This affects the ability of farmers to appropriately plan and manage their crop production pipelines and market processes. There is therefore a need to develop automated methods of crop yield estimation. However, the development of accurate machine-learning methods for crop yield estimation depends on the availability of appropriate datasets. There is a lack of such datasets, especially in sub-Saharan Africa. We present curated image datasets of coffee and cashew nuts acquired in Uganda during two crop harvest seasons. The datasets were collected over nine months, from September 2022 to May 2023. The data was collected using a high-resolution camera mounted on an Unmanned Aerial Vehicle . The datasets contain 3000 coffee and 3086 cashew nut images, constituting 6086 images. Annotated objects of interest in the coffee dataset consist of five classes namely: unripe, ripening, ripe, spoilt, and coffee_tree. Annotated objects of interest in the cashew nut dataset consist of six classes namely: tree, flower, premature, unripe, ripe, and spoilt. The datasets may be used for various machine-learning tasks including flowering intensity estimation, fruit maturity stage analysis, disease diagnosis, crop variety identification, and yield estimation.
ABSTRACT
The sweetpotato breeding process involves assessing different phenotypic traits, such as the sensory attributes, to decide which varieties to progress to the next stage during the breeding cycle. Sensory attributes like appearance, taste, colour and mealiness are important for consumer acceptability and adoption of new varieties. Therefore, measuring these sensory attributes is critical to inform the selection of varieties during breeding. Current methods using a trained human panel enable screening of different sweetpotato sensory attributes. Despite this, such methods are costly and time-consuming, leading to low throughput, which remains the biggest challenge for breeders. In this paper, we describe an approach to apply machine learning techniques with image-based analysis to predict flesh-colour and mealiness sweetpotato sensory attributes. The developed models can be used as high-throughput methods to augment existing approaches for the evaluation of flesh-colour and mealiness for different sweetpotato varieties. The work involved capturing images of boiled sweetpotato cross-sections using the DigiEye imaging system, data pre-processing for background elimination and feature extraction to develop machine learning models to predict the flesh-colour and mealiness sensory attributes of different sweetpotato varieties. For flesh-colour the trained Linear Regression and Random Forest Regression models attained R2 values of 0.92 and 0.87, respectively, against the ground truth values given by a human sensory panel. In contrast, the Random Forest Regressor and Gradient Boosting model attained R2 values of 0.85 and 0.80, respectively, for the prediction of mealiness. The performance of the models matched the desirable R2 threshold of 0.80 for acceptable comparability to the human sensory panel showing that this approach can be used for the prediction of these attributes with high accuracy. The machine learning models were deployed and tested by the sweetpotato breeding team at the International Potato Center in Uganda. This solution can automate and increase throughput for analysing flesh-colour and mealiness sweetpotato sensory attributes. Using machine learning tools for analysis can inform and quicken the selection of promising varieties that can be progressed for participatory evaluation during breeding cycles and potentially lead to increased chances of adoption of the varieties by consumers.