Liveweight estimation in ki̇lis goats with image processing and machine learning

The aim of this study is to predict the live weight of Kilis goat kids using image processing and machine learning techniques. The research was conducted at the Small Ruminant Unit of the Faculty of Agriculture, Dicle University, between July 2023 and March 2024. The study's material consisted of morphological measurements and back images obtained from 9 Kilis goat kids, which were collected regularly over a period of 10 months. The goats were selected based on similar age ranges within the Kilis breed. The back images were captured using a GoPro device with 4K resolution (3840x2160 pixels) and 30 frames per second (fps). These images were used to extract features for the dataset intended for machine learning. The images were initially labeled using the Labelme program in Python, with the labels being generated based on the references in the images. After the background removal of the back images, the resulting images were analyzed using ImageJ to obtain morphological measurements. Various machine learning models (Random Forest, SVR, MARS, and XGBoost) were employed to compare different datasets and test performance. The predicted performance of these models was evaluated based on statistics such as R² (Coefficient of Determination), MAE (Mean Absolute Error), and RMSE (Root Mean Squared Error) for the test data. The results indicated that the XGBoost model exhibited the best performance on the training dataset with R² = 0.98, MAE = 0.50, and RMSE = 0.64, but showed lower performance on the test dataset with R² = 0.01, MAE = 0.01, and RMSE = 5.36. The performance of the other models was as follows: Random Forest (R² = 0.85, MAE = 1.55, RMSE = 1.89), SVR (R² = 0.51, MAE = 2.26, RMSE = 3.19), and MARS (R² = 0.19, MAE = 3.10, RMSE = 3.91). This research demonstrates that image processing and machine learning techniques offer a reliable and practical alternative for predicting the live weight of small ruminants. Moreover, it represents a significant step towards the application of technological advancements in the livestock industry and the digital transformation of flock management. Keywords: Kilis Goat, Machine Learning, Image Processing, Live Weight prediction