Cattle Weight Estimation Using Fully and Weakly Supervised Segmentation from 2D Images

Weight information is important in cattle breeding because it can measure animal growth and be used to calculate the appropriate amount of daily feed. To estimate weight, we developed an image-based method that does not stress requires no manual labor. From a 2D image, mask was obtained by segmenting background, weights were estimated using deep neural network with residual connections extracting weight-related features from segmentation mask. Two image methods, fully weakly supervised segmentation, compared. The uses Mask R-CNN model learns ground truth generated labeling as correct answer. activation visualization map proposed this study. first creates more precise mask, but second require labeling. body weight statistical segmented region. In experiments, following performance results obtained: mean average error 17.31 kg absolute percentage 5.52% for 35.91 10.1% segmentation.