Towards on-farm pig face recognition using convolutional neural networks

Distinguishing Individual Red Pandas from Their Faces

“…As summarized in Table 1, automatic individual identification methods have been studied for a number of species, including African penguins [3], northeast tigers [4], cattle [5], lemurs [6], dairy cows [7], great white sharks [8], pandas [9], primates [10], pigs [11], and ringed seals [12]. Different species usually have largely different appearance; however, different individual animals of the same species may differ quite slightly in their appearance, and can be distinguished only by fine-grained detail.…”

“…For those species that have salient characteristics in their appearance (e.g., the spots on the breast of penguins [3], and the rings on the body of ringed seals [12]), individual identification can be done by extracting and comparing their salient features. For those species that have subtle appearance differences between different individuals, such as pigs [11], lemurs [6], and pandas [9], the most common solution to individual identification is to focus on the body parts with relatively rich textures and extract discriminative features from the parts. Red pandas obviously belong to those species that have subtle appearance differences between different individuals.…”

“…Compared with existing animal identification methods, ours is fully automatic. Almost all existing methods are based on pre-cropped pictures of specific body parts, such as the tailhead images of dairy cows [7] and face images of pig [11]. In contrast, our method takes the image of a red panda as input and automatically detect its face, extracts features and matches the features to the ones enrolled in the gallery to determine its identity.…”

“…Identification using microchips is generally more accurate, but as an invasive approach, it is hurtful to red pandas and unfriendly to operate. With the rapid development of computer vision technology in the past decade, some researchers attempt to automatically identify individuals of specific species based on images of the animals [3,4,5,6,7,8,9,10,11,12]. Following the pipeline of typical pattern recognition systems, they extract discriminative features from certain body parts of the animals, compute the similarity scores between images of the animals based on the extracted features, and finally determine the identities of the individual animals in the images according to their similarity scores with the reference images.…”

Giant Panda Face Recognition Using Small Dataset

“…Panda is one of them. To recognize animals for food safety (e.g., pig [2] and cow [3]), entertainment, sport (e.g., horse [4]) and wild animal conservation (e.g., chimpanzee [5]), biometric technology has been applied. More information about applying biometric technology to animals can be found in [1].…”