<![CDATA[In general, in cattle farming, individual identification is often done using text numbers on the cattle's ears, which is considered efficient. However, field observations reveal limitations of this technique, particularly in terms of data redundancy and new registration validation. Sometimes, errors can occur when cattle identity numbers are exchanged between the farm and the livestock market. Therefore, an intelligent biometric identification system attached to each cattle, such as the pattern on the muzzle, similar to human fingerprints, is needed. In this study, we collected and published primary cattle muzzle data as a dataset in a cloud repository. We also implemented the use of muzzle image data with a convolutional neural network algorithm in TensorFlow as a step for further development. The recognition implementation using the MobileNetV2 architecture resulted in an 83% accuracy rate for 30 individual cattle classes out of a total of 210 primary dataset divided into training and testing data.]]>
