For each of a plurality of images, a learning device acquires first training data including an image, a class of each of one or more objects shown in the image, and the number of objects, from a storage. The learning device trains a feature extraction part, a class estimation part, and an object number estimation part using the first training data. For each of images of which the number is smaller than the plurality of images, the learning device acquires second training data including an image and coordinates of each of one or more objects in the image, from the storage. The learning device causes the trained feature extraction part to output a feature map of the image included in the second training data, and trains a coordinate estimation part using the feature map and the coordinates included in the second training data.