In order to automatically detect the quantitative proportion of pigment deposition in the beaks of cephalopods, the deep learning network model of Mask-RCNN was used to realize the recognition and segmentation of beaks and its pigment deposition, and a new method of quantitatively automatic measurement for pigment deposition proportion in beaks was proposed. First of all, the beaks and their pigmentation were labeled, and the results were converted into a training set and imported into the residual network (Resnet50) to extract the characteristics of the beaks and their pigmentation. Based on the feature pyramid network (FPN), the features of each layer were merged, and then the region proposal network (RPN) was used to learn the features and generate candidate frames. Finally, the candidate frame was subjected to Non-Maximum Suppression (NMS) to obtain the candidate area of beaks and pigmentation, realizing the intelligent detection of the proportion of pigment deposition in beaks. The experiments showed that the segmentation accuracy of the upper beaks of the cephalopods was 93.60% and the pigmentation accuracy 92.47%; the segmentation accuracy of the lower beaks was 91.78% and the pigmentation accuracy 88.78%. The results show that the deep learning network model of Mask-RCNN could get the proportion of pigment deposition in the beaks and its pigment deposition, providing theoretical references for future studies in cephalopod feeding ecology.