simply udderly clever

The Silent Herdsman platform has its roots in the £4.75 million development programme funded by the ITI Techmedia with the goal of defining the market need and producing a prototype system for detecting the onset of oestrus (heat). Beyond the end of the programme, the Silent Herdsman has undergone significant development in terms of a robust solution and increased performance.

A number of key validation exercises have been undertaken to prove functionality and performance throughout the evolution of the product. These exercises have been executed through independent assessment of the accuracy of the heat detection under controlled environments and in comparison with other commercial systems offering heat detection. A number of farming locations have been utilised, carried out by experienced herdsman and animal scientists. The objective of the validations was to provide an independent comparison of the operational performance of the Silent Herdsman oestrus detection collar referenced against established ground methods that use truthing data derived from recordings of milk progesterone samples and other commercial systems. In all cases, predicted oestrus dates were independently compared with truthing data to yield the performance statistics.

Milk progesterone sampling records the daily levels of progesterone within the milk produced by individual cows and is accepted as being one of the most reliable indicators of oestrus. This method was used to determine the oestrus state of cows within the evaluations, delivering a benchmark of truth data to assess the performance of the Silent Herdsman system and its competitors. In all cases, milk progesterone samples were recorded by experienced staff and processed at a certified laboratory. The outputs of the processed samples were collated and presented in a format that allowed ready comparison with the oestrus prediction outputs from the collars. In cases where data was processed offline, the predictions were made on ‘blind’ data sets with a comparison to the truth data. Indications of an oestrus event within this period are recorded as a True Positive (TP). Oestrus predictions outside this period are recorded as a False Positive (FP). If no oestrus event is recorded within the oestrus period then a False Negative is recorded (FN) and if no oestrus predictions are made outside the oestrus period then a True Negative (TN) is recorded. Direct performance comparisons were made between the outputs of other commercial hat detection systems and to that of the Silent Herdsman systems respectively.

A range of performance metrics are used to quantify the performance of oestrus detection systems: