Body condition scoring (BCS) is a subjective manual method based on experience of individuals to assess fat reserves in the cow, of value to milk production, re-conception, regaining reserves and feeding. This is normally done at intervals to assist in monitoring loss and gain in body condition to ensure expected milk production, optimal health, conception and ease of calving. With increasing herd size the manual method is laborious which suggests that if an automated method can be developed which is quick, consistent and accurate, the method would benefit management in time, labour and genetic assessment. Such automated methods based on regression and Convolutional Neural Network (CNN) approaches have been reported in the literature. This has been taken further by the authors cited.

A common approach to automated BCS scoring is to utilise a CNN-based model trained with data from a depth camera. The approach of the authors was to make use of three depth cameras capturing depth frames of cows placed at different positions near the rear of the cow to train three independent CNNs. Ensemble modelling was then used to combine the estimations of the three individual CNN models.

Using the 5-point BCS scale, the results showed that BCS prediction by the three-camera ensemble was significantly improved from about 80% to 89%+, compared to the single depth camera and CNN model approach, which is highly satisfactory. The authors, however, noted in line with literature reports that it was difficult to accurately distinguish between BCS values lower than 2 or higher than 4, which may require further work.

Comment: BCSs of lower than 2 and higher than 4 should never occurr in commercial herds!