Abstract: A method by one or more computers for learning a machine learning model classifying an input image into known classes and an unknown class by extracting a feature of the input image. The method comprises acquiring a set of features of a plurality of samples of training data each of which is with a label, calculating a loss function based on the set of the features, and learning the machine learning model to reduce the loss function. The loss function is calculated to include a first loss function and a second loss function. The first loss function is configured to become larger as first distances become smaller than a predetermined margin. The second loss function is configured to become larger as second distances become larger and as third distances become smaller.

Abstract: An image recognition method according to one aspect of the present invention acquires a camera image generated by capturing a subject using a camera (three-dimensional sensor). A plurality of coordinates corresponding to a plurality of pixels included in a predetermined area in the camera image are acquired. Subject distance information indicating a distance from the subject to the camera in the plurality of pixels is acquired. Then the posture of the subject surface included in the subject in the predetermined area is estimated based on the plurality of coordinates and the plurality of pieces of subject distance information that have been acquired.

Abstract: A computer-based method/system of dynamic category object recognition for estimating pose and/or positioning of target objects and target object's parts. The method/system may recognize a target object and the target object's parts. The method/system may segment and extract data corresponding to the target object and the target object's parts, and estimate the pose and positioning of the target object and the target object's parts using a plurality of stored object models. The dynamic method/system may supplement or modify the parameters of the plurality of stored object models and/or store learned object models. The learned object models assist in recognizing and estimating pose and/or positioning of newly encountered objects more accurately and with fewer processing steps. The method and system may include a processor, a sensor, an external device, a communications unit, and a database.

Abstract: An image recognition method according to one exemplary aspect of the present invention including the steps of: acquiring a shooting image generated by capturing an image of an object using an image generating device; acquiring subject distance information indicating a distance from the object to the image generating device at a target pixel in the shooting image; extracting an image pattern corresponding to the acquired subject distance information from a plurality of image patterns which are created for detecting one detection object in advance and are associated with the different distance information, respectively, and performing a pattern matching using the extracted image pattern against the shooting image.

Abstract: A computer-based method/system of dynamic category object recognition for estimating pose and/or positioning of target objects and target object's parts. The method/system may recognize a target object and the target object's parts. The method/system may segment and extract data corresponding to the target object and the target object's parts, and estimate the pose and positioning of the target object and the target object's parts using a plurality of stored object models. The dynamic method/system may supplement or modify the parameters of the plurality of stored object models and/or store learned object models. The learned object models assist in recognizing and estimating pose and/or positioning of newly encountered objects more accurately and with fewer processing steps. The method and system may include a processor, a sensor, an external device, a communications unit, and a database.