Method and apparatus for determining object posture in image, device, and storage medium

A method for determining a posture of a target object in an image, comprising:

• obtaining a standard image of the target object, and a plurality of distorted images of the target object;
• inputting the standard image and the plurality of distorted images into a convolutional neural network for training to obtain a training model parameter of the convolutional neural network;
• and
• sending, by processing circuitry, the training model parameter to a terminal device, wherein the training model parameter is configured to cause the terminal device to:
• obtain a real-time image of the target object,
• identify at least one first image block from the real-time image, the first image block being a local image of the real-time image,
• determine, according to the training model parameter, a label image block matching the at least one first image block, the label image block being a local image of the standard image of the target object, and
• determine the posture of the target object according to the at least one first image block and the label image block,
• wherein in the identification of the at least one first image block, (i) a plurality of local features on the real-time image are detected, and (ii) one of the plurality of local features that is determined not to be at an edge of the real-time image is determined to be a first image block of the at least one first image block.

An apparatus for determining a posture of a target object in an image, comprising:

• processing circuitry configured to
• obtain a training model parameter of a convolutional neural network of the target object from a server,
• obtain a real-time image of the target object,
• identify at least one first image block from the real-time image, the first image block being a local image of the real-time image,
• determine, according to the training model parameter, a label image block matching the at least one first image block, the label image block being a local image of a standard image of the target object, and
• determine the posture of the target object according to at least one first image block and the determined label image block,
• wherein the processing circuitry is further configured to
• detect a plurality of local features on the real-time image, and
• determine one of the plurality of local features that is determined not to be at an edge of the real-time image to be a first image block of the at least one first image block.

An apparatus for determining a posture of a target object in an image, comprising:

• processing circuitry configured to
• obtain a standard image of the target object, and a plurality of distorted images of the target object,
• input the standard image and the plurality of distorted images into a convolutional neural network for training to obtain a training model parameter of the convolutional neural network, and
• send the training model parameter to a terminal device, wherein the training model parameter is configured to cause the terminal device to
• obtain a real-time image of the target object,
• identify at least one first image block from the real-time image, the first image block being a local image of the real-time image,
• determine, according to the training model parameter, a label image block matching the at least one first image block, the label image block being a local image of the standard image of the target object, and
• determine the posture of the target object according to the at least one first image block and the label image block,
• wherein in the identification of the at least one first image block, (i) a plurality of local features on the real-time image are detected, and (ii) one of the plurality of local features that is determined not to be at an edge of the real-time image is determined to be a first image block of the at least one first image block.