Open Set Sheep Face Recognition Based on Euclidean Space Metric

Figure 9

Triplet loss: negative refers to the negative sample most similar to the sample, and positive refers to the positive sample least identical to the sample. Through the supervision of the triple loss function and combined with the “semihard” triple data training network, the neural network can learn the ability to distinguish positive and negative samples better. (a) Before learning, the model cannot correctly measure the spatial distance between the standard sample and the positive sample and the negative sample. (b) The model’s learning process of “semihard” data optimizes the model’s ability to measure the spatial distance between the sample and the positive and negative samples. (c) It means that after learning the triple loss function supervised model, the model can correctly measure the spatial distance between standard samples and positive and negative samples and accurately distinguish sample categories.